Find the right support

Research date: September 15, 2026.

Every figure in this chapter was checked against the best source available on or immediately before that date. Accelerator terms, batch dates, deadlines, credit amounts and eligibility rules change frequently — several of the programs below changed terms within the last twelve months, and at least one major one restructured itself out of half its programs. Nothing here should be treated as current at the moment you read it. Check the official page before you apply, and treat any secondary source (including this one) as a starting point.

This chapter is a companion to the funding-instruments chapter, which covers SAFEs, convertible notes, priced rounds, valuation caps, dilution mathematics, and the mechanics of grant disbursement. Here the subject is the institutions themselves — who they are, what they actually offer, who they are right for, and whether the exchange is worth it.

How to read the evidence in this chapter

Accelerator research has a specific and severe evidence problem, and it is worth being explicit about it before any numbers appear.

Labels used throughout:

• [Verified] — Confirmed on the program's own official page or in a primary document (a press release from the organization, a government program page, a regulatory filing), on or shortly before September 15, 2026.
• [Reported] — Published by credible journalism or a named third party, but not confirmed on the program's own site.
• [Unverified] — A figure that circulates widely but which I could not confirm on a primary source at this date, or which appears only on secondary aggregator sites of unknown quality. Several of these are probably stale.
• [Analysis] — My interpretation, clearly separated from sourced fact.

Four standing conflicts of interest apply to almost every number in Part A:

1. Accelerators publish their own success statistics, and there is no audit. "Our alumni have raised $X billion" is a marketing claim computed by the party being marketed, and the denominator is almost never disclosed. A program can truthfully claim "$5B raised by alumni" when that sum was raised by four companies out of seven hundred.
2. Survivorship bias is total, not partial. Portfolio pages list companies that still exist; dead companies quietly disappear from the website. Every "alumni outcome" figure is computed over survivors.
3. Selection versus treatment is unresolved and mostly unresolvable. Does the accelerator cause the outcome or select companies that were going to do well anyway? For the top programs, nobody knows the split. Part B covers what the academic literature can and cannot establish.
4. A large content-marketing industry exists around accelerator applications. Many pages ranking for "YC acceptance rate 2026" are produced by application-coaching services or SEO affiliates with an interest in making programs seem both prestigious and gettable. Where a figure appears only on such sources, I mark it Unverified.

A fifth point, specific to 2026: capital concentration in AI has changed what an accelerator is for — developed in Part B, but it colors every term sheet in Part A.

Directory field schema

Each program below uses the same fields so the directory can be mapped cleanly to a structured database:

PART A — THE PROGRAM DIRECTORY

A1. The top-tier accelerators

Y Combinator

Type: Accelerator Stage: Idea stage through early revenue; increasingly pre-product Geography: San Francisco (in-person required) Verification status: Deal terms and application process verified on ycombinator.com, September 2026.

Capital & equity [Verified]. YC's standard deal is $500,000 total, split across two instruments: $125,000 on a post-money SAFE for 7% of the company, plus $375,000 on an uncapped SAFE with a Most Favored Nation (MFN) provision (Y Combinator, "The YC Deal", accessed September 2026). The uncapped MFN portion converts at whatever the best terms are in the next priced round — YC's own worked example is that at a $15M post-money next round, the $375K becomes roughly 2.5%. YC takes pro rata rights to maintain ownership in later rounds and charges no fees.

The structure has been stable since the $500K deal was introduced in January 2022 (Y Combinator, "YC's $500,000 Standard Deal"). I found no evidence of a change to the headline terms in 2025 or 2026 — the deal page as of September 2026 states the same $125K/7% + $375K/MFN structure. Anyone who tells you the terms changed this year should be asked for a link to ycombinator.com.

The dilution math is frequently misunderstood: 7% is the fixed floor, not the total. Total YC dilution is 7% plus whatever the $375K converts into, inversely proportional to how well you raise — roughly 8.25% if the next round is at $30M post, roughly 11.7% if it is at $8M post. The uncapped MFN structure rewards companies that raise high, but it is not free, and "YC takes 7%" is an incomplete description.

Cadence & timing [Verified]. This is the most consequential recent change and the one most commonly reported wrong. YC now runs four batches a year — Winter, Spring, Summer and Fall — having moved from two to four cohorts. As of September 2026, applications are open for the Winter 2027 batch, with an on-time deadline of November 2 at 8pm PT and decisions by December 11; the Winter 2027 program runs January to March in San Francisco, opening with a mandatory three-day in-person kickoff (Y Combinator, "Apply", accessed September 2026). YC also offers an Early Decision option for later batches. Critically, YC invests as soon as a company is accepted and does not wait for the batch to start — which means acceptance is a liquidity event, not just a calendar event.

Application process [Verified]. Online application at apply.ycombinator.com; promising applicants get a video interview (for W27, scheduled November–December); decisions are typically made the same day as the interview, and YC gives detailed feedback either way. The interview is famously short — roughly ten minutes — and the application text does most of the filtering.

Batch size [Reported]. A third-party analysis of YC's own company directory counts 593 companies across 2024's batches (Summer 249, Winter 251, Fall 93) and 632 across 2025's four batches (roughly 145–168 each) (Ellenox, "Y Combinator Statistics and Insights," February 2026). This is scraped data, not official, and YC does not publish batch sizes as a policy. The direction is clear and useful though: YC broke the pandemic-era mega-batch (Winter 2022 peaked near 400 companies) into four cohorts of roughly 150, so total annual throughput is similar but the per-batch experience is smaller. [Analysis] That is a meaningful quality change — a 150-person batch has denser partner attention than a 400-person one — and it is underappreciated by founders comparing YC to its 2021 reputation.

Acceptance rate [Unverified]. Widely cited figures range from under 1% to about 1.5–2%. One frequently-linked estimate puts it at 1.5–2% from roughly 15,000 applications per batch (ValueAdd VC, 2026); others cite ~1%. YC does not publish this number, and every source claiming it is estimating. All such sources are aggregators or application-coaching adjacent. Treat "roughly one to two percent" as the honest resolution and ignore the decimal points.

Mentorship model. Group office hours with a YC partner plus individual office hours on request; weekly dinners with speakers; an internal forum (Bookface) where alumni answer questions; Work at a Startup for hiring. The structure is deliberately low-touch relative to programs like Techstars — YC's partners are a scarce resource spread across ~150 companies per batch. [Analysis] The mentorship is good but the honest characterization is that YC sells accountability, peer pressure and a deadline more than it sells advice. The weekly growth-number ritual is the product.

Network value. This is the genuine, hard-to-replicate asset and the main reason the 7% is defensible. It has three parts: (1) the alumni network and Bookface, which functions as a high-signal question-answering system and a customer base — YC companies sell to each other at meaningful volume; (2) the investor-facing signal, which materially lowers the cost of raising a seed round; (3) Demo Day, which compresses fundraising into a competitive, time-boxed process. [Analysis] Of these, the investor signal is the most valuable and the most clearly causal, because it operates on how third parties behave rather than on anything the company does internally.

Follow-on record. YC's own public claim is that its portfolio's combined valuation exceeds several hundred billion dollars, driven by a handful of outliers (Airbnb, Stripe, DoorDash, Coinbase, Instacart, Dropbox, Reddit). Scraped third-party counts put the portfolio above 5,600 companies with a combined valuation figure above $600B (Ellenox, February 2026). This number is close to meaningless as a guide to an individual founder's expected outcome. A portfolio valuation dominated by Stripe tells you about Stripe. The relevant statistic — median outcome for a YC company, or the fraction that reach a Series A, or the fraction that return capital — is not published by YC and is not reliably available elsewhere.

Best fit. Venture-track software founders who (a) intend to raise institutional capital within 6–12 months, (b) can relocate to San Francisco, (c) are early enough that 7%+ is cheap relative to the signal gained, and (d) do not already have strong investor access. Founders who already have warm access to good seed investors get the least incremental value from YC, because the signal is the largest component of what they are buying.

Drawbacks. The equity is real and permanent, and it sits at the earliest and therefore most expensive point in the cap table. The batch has homogenizing pressure — companies converge on similar metrics, similar pitches, similar B2B SaaS shapes, and Demo Day rewards legibility over originality. The four-batch cadence means a larger absolute number of YC companies in market each year, which modestly dilutes the scarcity of the signal. And selection effects are doing an enormous amount of the work — see Part B.

Techstars

Type: Accelerator (multi-program network) Stage: Pre-seed to seed Geography: Multiple cities globally; program-specific Verification status: Investment terms verified on techstars.com (announcement dated April 17, 2025). Program list verified September 2026.

The restructuring is the headline, and it matters. Between 2023 and 2025 Techstars went through the most significant contraction of any major accelerator brand. In February 2024 it announced it was moving its headquarters out of Colorado and ending its Boulder accelerator — the program in the city where the organization was founded (Colorado Sun, February 2024). It simultaneously shut down Techstars Seattle, saying it was shifting focus to cities with more venture activity (GeekWire, 2024). It also ended its JPMorgan-backed Washington DC-area accelerator (Technical.ly). CEO Maëlle Gavet publicly defended the changes, arguing a physical presence in a city is not necessary to invest there (TechCrunch, February 2024). A Boulder program was later revived in a different form (Colorado Sun, October 2024), and a Techstars Boulder program appears on the current program list as of September 2026 (Techstars, "Accelerators").

[Analysis] The strategic read: Techstars' old model — many geographically distributed programs, often underwritten by corporate or civic sponsors, producing high volume at modest average quality — stopped working when venture capital concentrated geographically and by sector. The organization is now smaller, more concentrated, and more dependent on corporate partners. The practical consequence is that "Techstars" is not one thing and its brand value varies enormously by program. Evaluate the specific program and the specific Managing Director, not the logo.

Capital & equity [Verified]. Techstars' current terms, announced April 17, 2025, are $220,000 total: a $20,000 Convertible Equity Agreement that converts to 5% common stock on a priced round of at least $1M, plus a $200,000 uncapped MFN SAFE that converts on the same round at the best terms offered to other investors (Techstars, "Investment Terms Update"). There is a side letter with pro rata, drag-along and information rights. No enrollment fee. US incorporation (or a foreign equivalent) is required before investment. Asia-Pacific programs use a $100,000 uncapped MFN SAFE instead of $200,000.

Two things are unusual. First, the 5% is in common stock, not preferred — Techstars sits alongside founders rather than ahead of them in a liquidation, which is genuinely founder-aligned. Second, the deal has improved over time: pre-2024 it was widely described as $20K for 6% plus a $100K note; a 2024 update moved to $120,000 (Techstars, 2024 terms); April 2025 raised it to $220,000 while dropping fixed equity to 5%. [Analysis] More money, less fixed equity, more of the total uncapped — a direct competitive response to YC and to founders' improved outside options.

Cadence & timing. Programs run three months, mentorship-driven, with per-program application windows. As of September 2026 the accelerator list includes roughly 16 named programs, among them Techstars Boston, Boulder, Chicago, London, New York City, Tokyo, the Anywhere Accelerator (remote), Techstars Space, Techstars AI Health Baltimore, Techstars Alabama EnergyTech, USC and Techstars, Techstars Columbus with Ohio State, plus corporate-partner programs with ABN AMRO, Northwestern Medicine, Ecolab and Permanente Medicine (Techstars, "Accelerators", accessed September 2026). Deadlines are program-by-program; there is no single Techstars deadline.

Mentorship model. This is genuinely Techstars' differentiator and where it beats YC on paper. "Mentor madness" — a structured, high-volume set of mentor meetings in the first weeks, from which founders select a small number of lead mentors — is a real, well-designed mechanism, and the mentor-to-company ratio is far higher than YC's. The quality is entirely dependent on the local mentor pool, which is exactly why the geographic contraction matters.

Network value. Large alumni network; strong per-city operator networks in the remaining strong programs; corporate-partner programs offer something YC cannot — a named enterprise willing to pilot. For a company selling into healthcare, energy, or financial services, a corporate-sponsored Techstars program can deliver a design partner, which is worth more than a Demo Day.

Follow-on record [Unverified]. Techstars publishes aggregate portfolio figures. As with all accelerators, these are self-published, unaudited and survivorship-biased. The more useful signal is program-specific: ask a Managing Director directly what fraction of the last three cohorts raised a priced seed round within 18 months, and treat an unwillingness to answer as an answer.

Best fit. Founders with a specific corporate-partner or sector fit (space, energy, health systems, fintech-with-a-bank); founders outside the Bay Area who need a structured mentor network more than they need a Demo Day; founders for whom the higher-touch model beats YC's volume model. Also: founders who did not get into YC and are choosing between Techstars and no accelerator — for whom the honest answer is often "it depends entirely on which program."

Drawbacks. Brand variance across programs is the central problem. A weak Techstars program costs 5% common plus three months for a mentor network that may not be worth it. The MFN structure means total dilution is again more than the headline 5%. And the organization's own recent instability is a legitimate consideration — the program you join may not exist in three years, which affects the durability of the alumni network you are buying.

500 Global (formerly 500 Startups)

Type: Accelerator + global VC Stage: Pre-seed/seed with early traction Geography: Silicon Valley (in-person, four months) Verification status: Terms stated on 500.co, but the page I retrieved carried a 2024 date stamp and an expired deadline. Treat as probably-current but not confirmed for 2026.

Capital & equity [Reported]. "$150,000 investment for a 6% stake, subject to terms and diligence" for the Flagship Accelerator (500 Global, "Flagship Accelerator"). Historically 500 also charged a program fee taken out of the investment — a structure that made the effective economics worse than the headline. I could not verify on the official page whether a fee still applies in 2026, and founders should ask directly.

Cadence & timing [Unverified]. Four-month in-person program in Silicon Valley; batch numbering was in the mid-30s at last public reference. The deadline shown on the official page when accessed in September 2026 was stale, which is itself a mild signal about the program's current activity level.

Mentorship & network. 500's historical strength was growth marketing and distribution, and an unusually international portfolio — it has been one of the most globally distributed accelerator networks, with meaningful presence and separate funds across Southeast Asia, the Middle East, and Latin America. For a founder from an emerging market who wants a US-legible investor network, this has been a real and differentiated asset.

Follow-on record [Unverified]. 500 publishes large aggregate numbers across a portfolio of several thousand companies. As with all such figures: self-published, unaudited, survivorship-biased, and dominated by a handful of outliers.

Best fit. International founders seeking US market entry and US investor legibility; growth-stage-adjacent companies with a product and early revenue that need distribution help more than product help.

Drawbacks. 6% for $150K implies a $2.5M post-money valuation — meaningfully more expensive per dollar than YC's blended terms, without the equivalent signal. [Analysis] 500 Global is the clearest case in the top tier where the price no longer obviously matches the brand. Its strongest era was 2012–2018; its current brand strength among US seed investors is well below YC's and arguably below a strong Techstars program's. If you are a US founder with US investor access, the case is weak. If you are a founder in Jakarta or Riyadh trying to become legible to Sand Hill Road, it is much stronger.

Antler

Type: Pre-team venture "residency" + global VC Stage: Pre-idea and pre-team — earlier than any other program here except EF Geography: 26+ locations globally; US programs in San Francisco, New York, Austin Verification status: Program structure verified on antler.co September 2026; specific 2026 investment terms not verified — see below.

The model. Antler is not an accelerator in the YC sense. It admits individuals, not companies, and runs a structured co-founder matching and idea-formation process. You arrive without a team and often without an idea; you spend weeks meeting other residents, forming and dissolving teams, and testing concepts; at the end, teams pitch an investment committee, and a minority receive a term sheet.

Capital & equity [Partially verified]. The current Antler US page states an initial commitment of $500K–$1M for selected founders and describes a residency of up to three months across San Francisco, New York City and Austin (Antler, "Antler in the US", accessed September 2026). The equity percentage is not stated on that page. The most recent specific terms I could verify are from October 2023: $250,000 at a $2.75M post-money valuation (9.09%), plus a $2,500 relocation grant (Built In Austin, on Antler's updated US terms). Those 2023 figures should be treated as historical. The 2026 headline commitment is larger, but without a stated valuation the dilution is unknown. Ask Antler directly for the current post-money and the split between initial check and reserved follow-on — "$500K–$1M commitment" is a range that includes money you may never receive.

Some regional programs advertise their own improved terms independently — Antler Singapore, for example, announced separate terms and a six-week residency (Antler Singapore). Antler's terms are location-specific. There is no single Antler deal.

Eligibility [Verified]. For the US programs, at least one co-founder must have unrestricted US work authorization; Antler does not sponsor visas for US residencies.

Best fit. Experienced operators who want to found a company, have no co-founder, and have no strong idea — a genuinely underserved profile. Also strong for founders in markets where the local ecosystem has no equivalent structure (Antler's geographic breadth is its real differentiator; it operates in cities no US accelerator touches).

Drawbacks. [Analysis] Three real ones. First, dilution is high relative to what the company is — a team formed six weeks ago giving up close to 10%, at the very bottom of the cap table, compounds through every subsequent round. Second, co-founder matching under time pressure is a structurally risky way to pick a co-founder; founder conflict is the most common cause of early-stage failure, and a relationship formed in six weeks under an incentive to form some team is a worse selection process than years of working together. Third, acceptance to the residency is not acceptance to funding — a large fraction of residents leave with nothing after three months. The program says so; applicants consistently under-weight it.

Entrepreneur First (EF)

Type: Pre-team "talent investor" Stage: Pre-idea, pre-team Geography: Hub cities across North America, Europe and Asia (London, Bangalore, others); second phase in San Francisco Verification status: Structure verified on joinef.com September 2026.

The model. EF invented the "talent investing" category that Antler later scaled. EF explicitly states you do not need a co-founder, an idea, or even a technical background — the requirement is demonstrating "exceptional outcomes relative to your peers" (Entrepreneur First, FAQs, accessed September 2026).

Structure [Verified]. Two 12-week phases. FORM runs at a local hub and covers co-founder matching and idea formation, supported by a Talent Investment grant — equity-free, covering living costs, not repayable. Teams that pass the Investment Committee move to LAUNCH in San Francisco, with up to $250,000 in equity investment and over $600,000 in partner credits (AWS, Stripe, OpenAI and others). Full-time and in-person is required. Applications are reviewed on a rolling basis.

[Unverified] EF's equity percentage for the LAUNCH investment is not stated on the public FAQ; historically EF has taken roughly 8–10% at this stage, but I could not verify a current figure on an official source and will not repeat a stale one. Terms vary by hub, and EF's FAQ says so explicitly.

Best fit. Deep-technical people — PhDs, research engineers, strong ICs from large tech companies — who are highly capable and have no company and no co-founder. EF has historically been the best program in the world at this specific job, and its European track record (Tractable, Cleo, Omnipresent and others) is real. The equity-free grant during FORM is a materially better deal than Antler's structure for the pre-investment phase, because you are not diluting while you search.

Drawbacks. Same structural co-founder-matching risk as Antler. The relocation requirement to San Francisco for LAUNCH is a hard gate for many international participants (visas are the binding constraint, not ambition). And like Antler, completing FORM does not guarantee investment. [Analysis] EF's honest value proposition is: if you are extremely talented and have no path into founding, this is the best available path, and you should expect to pay for it in dilution and optionality.

AlchemistX / Alchemist Accelerator

Type: Accelerator, enterprise/B2B and deep tech specialist Stage: Seed Geography: San Francisco flagship; programs in Chicago, Doha, Japan Verification status: Program structure verified on alchemistaccelerator.com September 2026; financial terms not published on the official site.

Structure [Verified]. Six-month in-person flagship program in San Francisco for enterprise and deep-technology founders, with weekly coaching, curated investor introductions, and a private investor showcase rather than a public demo day. Regional programs run in Chicago, Doha and Japan (Alchemist, "Programs", accessed September 2026). Alchemist's self-published track record is 750+ startups graduated and $5B+ raised by portfolio companies — self-reported, unaudited, survivorship-biased.

Capital & equity [Unverified]. Alchemist does not publish investment amount or equity on its programs page. Historically the deal has been described as a modest five-figure investment for mid-single-digit equity. Get the number in writing before applying; an accelerator that does not publish its terms is not automatically bad, but it is a reason to ask early.

The "AlchemistX" distinction. AlchemistX is Alchemist's corporate-innovation arm — it runs accelerator programs on behalf of corporations and governments rather than investing its own money in a general cohort. If a program is branded AlchemistX rather than Alchemist Accelerator, you are in a corporate or government-sponsored program with different economics, usually no equity taken and no investment made. This distinction is frequently blurred in third-party write-ups and matters a great deal.

Best fit. Enterprise-software and deep-tech founders whose main bottleneck is getting a first enterprise customer, not getting a seed round. Alchemist's differentiator is the mentor pool's density of enterprise-sales operators, which is genuinely rare among accelerators. The six-month length also fits enterprise sales cycles better than a 12-week program does.

Drawbacks. Lower investor-signal value than YC. Six months in person is a large time commitment. Undisclosed terms.

HF0 ("the residency for repeat founders")

Type: Residency Stage: Varies — typically founders with prior exits or prior scale experience Geography: San Francisco Verification status: Program philosophy and outcome claims verified on hf0.com September 2026; financial terms not published on the official site.

The model [Verified]. HF0 states plainly: "We only back 10 teams at a time" and positions itself as "the residency for repeat founders" (HF0, "Facts", accessed September 2026). Founders live and work together in a converted monastery in San Francisco for the duration of the batch, with the explicit design goal of removing every distraction from building. Recent batch start dates referenced on the site are September 13 and January 4, with demo days in December.

Self-reported outcomes [Verified as claims, not as audited facts]. HF0's own facts page claims: S25 teams averaged $82M valuation at demo day, with the top team at $20M annualized revenue; in W25 and F24, four of ten teams passed $3M+ revenue; S24 included four repeat unicorn founders. Notable alumni include Krea and Crossmint (HF0, "Facts").

[Analysis] These are the most impressive raw numbers of any accelerator in this chapter, and they are also the clearest illustration of selection effects in the entire industry. HF0 admits ten teams, several of whom are repeat unicorn founders. A repeat unicorn founder would raise at a high valuation regardless of whether they spent twelve weeks in a monastery. HF0's numbers are a measure of who HF0 admits, almost entirely. That does not make HF0 bad — concentrated attention on ten teams is a real product, and the founders who go clearly value the environment — but "HF0 companies average $82M valuations" is not a claim about what HF0 does to a company.

Capital & equity [Unverified]. Not published on the official site. Third-party sources circulate figures in the range of a several-hundred-thousand-dollar investment on an uncapped SAFE; I could not verify any specific number on a primary source and will not repeat one.

Best fit. Second- or third-time founders with a strong technical team who want extreme focus and a very small, very high-caliber peer group. First-time founders are explicitly not the target.

Drawbacks. Ten slots. The bar is a track record you either have or do not. Undisclosed terms.

South Park Commons (SPC) Founder Fellowship

Type: Fellowship / pre-idea community Stage: "Minus one to zero" — explicitly before an idea exists Geography: San Francisco, New York City, or Bangalore for the bootcamp; flexible after Verification status: Fully verified on southparkcommons.com, September 2026 — one of the most transparent programs in this chapter.

Capital & equity [Verified]. $1M total: $400K for 7% upfront, plus $600K guaranteed in your next external round, plus up to $1M in partner credits including Anthropic and OpenAI (South Park Commons, "SPC Founder Fellowship, Fall 2026").

[Analysis] The $600K guaranteed follow-on is the structurally interesting part. A guaranteed commitment into your next round removes seed risk at the moment you most need it, and is a stronger commitment than the "pro rata rights" most accelerators take — which are an option for them, not an obligation. On pure terms, SPC's $400K/7% + $600K guaranteed is the most founder-favorable headline deal among the pre-idea programs, and compares well to YC's $125K/7% + $375K MFN: more upfront cash for the same fixed equity.

Structure & timing [Verified]. Funding on signing; then an 8-week in-person bootcamp in a small cohort; then an open-ended residency "as long as you need" before raising a seed round. For the Fall 2026 cohort: applications closed August 2, 2026; interview notifications by August 30; bootcamp late September to late November 2026. Solo founders are welcome. No finalized idea required — the application asks how you generate ideas.

Best fit. Strong technical operators between things, who want funding and a serious peer community while they figure out what to build, and who want more upfront cash than YC provides. SPC's community (it began as a physical space and intellectual community in San Francisco, not as an accelerator) is unusually high-quality and unusually non-transactional.

Drawbacks. The 7% is charged before an idea exists, which is the maximum-uncertainty moment to price a company. The open-ended residency is a double-edged design: no deadline means no forcing function, and some fellows drift. The investor signal is real but narrower than YC's.

Neo

Type: Residency / accelerator Stage: Early, with a separate student track Geography: San Francisco (Jackson Square), plus a two-week Oregon bootcamp Verification status: Terms reported by TechCrunch, February 2026. New program — first cohort summer 2026 — so there is no track record for this structure yet.

Capital & equity [Reported]. Neo, founded by Ali Partovi, announced a residency investing $750,000 on an uncapped SAFE into cohorts of 12–15 startups, capped at 20 teams, two cohorts a year. Because the SAFE is uncapped, dilution is valuation-dependent: roughly 5% if the next round is at $15M, roughly 0.75% if it is at $100M (TechCrunch, February 19, 2026). Structure: a three-month San Francisco residency plus a two-week bootcamp in the Oregon mountains, with roughly 30 mentors. Neo also gives $40,000 no-strings grants to 5–8 college student teams to take a semester off.

[Analysis] This is the most aggressive competitive move against YC's terms in years, and it is worth understanding why it is possible. An uncapped SAFE with no fixed floor means Neo only does well if the company raises at a high valuation — the accelerator is betting on the top of its own distribution rather than taking a guaranteed slice. Partovi's stated rationale is confidence in picking future founders. The risk to founders is not the terms; it is that a program with no track record under this structure is selling an unproven network. Neo's earlier fellowship/scholars programs have a genuine reputation for identifying strong young technical talent, but the residency is new. The terms are excellent. The question is whether the rest is.

Best fit. Technically exceptional founders, especially young ones, who believe they will raise at a high valuation and therefore want uncapped rather than fixed dilution — the structure is precisely optimized for founders who expect to do well.

Drawbacks. Unproven program structure; small cohort; no follow-on record yet; the value of the network is a bet on Partovi's personal reach rather than on an institutional asset.

a16z Speedrun

Type: Accelerator run by a large VC firm Stage: Pre-seed to seed Geography: San Francisco (12 weeks) Verification status: Program scale verified on a16z.com September 2026; terms reported by TechCrunch, February 2026, not confirmed on a16z's own page.

Capital & equity. a16z's own page states up to $1M per startup over a 12-week program, with the next cohort (SR008) starting early 2027 in San Francisco (a16z, "Speedrun", accessed September 2026). TechCrunch, in the course of reporting Neo's terms, described Speedrun's structure as $500,000 for 10% via SAFE, with an additional $500,000 if a Series A closes within 18 months (TechCrunch, February 2026). These are compatible — "up to $1M" is the two tranches combined — but the 10% figure is [Reported], not verified on a16z's own site.

Scale [Verified]. Since launching in 2023, Speedrun has deployed over $180M across more than 200 startups, with a community of 600+ founders. The program provides operational support in finance, recruiting, GTM, HR, marketing, legal/regulatory and visas.

Best fit. Founders in a16z's thesis areas — Speedrun began with a games and interactive-media focus and broadened toward AI and consumer — who want a direct line into one of the largest venture firms and value visa support and operational scaffolding.

Drawbacks. [Analysis] 10% for $500K is the most expensive headline deal in the top tier — roughly a $5M post-money, which is not obviously bad on its own, but is a lot of fixed equity. More importantly, there is a structural consideration founders consistently underrate: an accelerator run by a single large VC firm creates signaling risk. If a16z passes on your Series A, every other firm knows a16z had twelve weeks of inside access and declined. YC, Techstars and SPC do not lead your next round, so their non-participation says nothing. Speedrun's involvement is high-value if a16z leads your next round and a measurable liability if it does not. Weigh that honestly against the operational support.

A2. Sector-specific and deep-tech programs

SOSV — HAX and IndieBio (now SOSV SF / SOSV NY)

Type: Deep-tech accelerator + VC Stage: Pre-seed, frequently pre-company Geography: New York and San Francisco (biology/life sciences); HAX is hardware-focused Verification status: Structure and application process verified on sosv.com September 2026; specific investment terms not published on the public FAQ.

A naming change that matters [Verified]. At the start of 2026, SOSV rebranded IndieBio to SOSV NY and SOSV SF to reflect an expanded scope beyond biotech (IndieBio / SOSV, accessed September 2026). Self-reported track record: 310 graduates who have collectively raised $3.6B, across food, agriculture, industrial decarbonization, AI, therapeutics and diagnostics. Self-published and survivorship-biased, as always — but note this is one of the few programs that reports a denominator (310), which makes the claim more interpretable than most.

Eligibility and process [Verified]. Rolling applications through sosv.com/apply. A minimum of two co-founders is required to receive funding — solo founders may apply but must find a co-founder before acceptance. Idea-stage is acceptable if "the technology [is] worked out at least in principle." International applicants are welcome, though US incorporation is recommended and SOSV assists with visas post-acceptance. Founders may apply to both HAX and the SOSV NY/SF biology programs. SOSV states explicitly that its program investment terms are non-negotiable (SOSV, "Application FAQs").

What makes it genuinely different. SOSV provides wet-lab space and equipment (for the bio programs) and hardware prototyping infrastructure (HAX). For a biology or hardware founder, this is not a soft benefit — it is the difference between needing $2M of capex and needing none. No software accelerator offers anything comparable in kind. [Analysis] This is the clearest case in the chapter where an accelerator's value is physical and causal rather than reputational: access to a BSL-2 lab you could not otherwise afford changes what experiments you can run, full stop.

Best fit. Scientist-founders in biology, materials, food/ag, climate hardware and adjacent fields at the point where they need lab access and a first check simultaneously.

Drawbacks. Deep-tech accelerator terms are typically more dilutive than software terms, reflecting the higher capital intensity and longer timelines; SOSV does not publish them, and says they are non-negotiable, so you learn them late in the process. Two-founder requirement excludes solo scientist-founders.

Activate (including Cyclotron Road)

Type: Fellowship — zero equity Stage: Pre-company; scientists and engineers with a hard-tech concept Geography: Berkeley (with Cyclotron Road at Lawrence Berkeley National Laboratory), Boston, Houston, New York, plus Activate Anywhere (nationwide, NSF-sponsored); a Singapore fellowship launching 2026 Verification status: Fully verified on activate.org, September 2026.

Terms [Verified]. More than $300,000 over two years, covering an annual living stipend, R&D costs and other expenses. The program takes zero equity (Activate, "The Fellowship", accessed September 2026). Two-year duration. Fellows get 1:1 mentorship, workshops, and access to national-lab facilities and networks. Activate welcomed 50 new fellows to its 2026 cohort (Activate, "Activate Welcomes 50 New Fellows to Cohort 2026").

[Analysis] For an eligible hard-tech scientist-founder, Activate is close to a dominant strategy and it is not close. Two years of salary plus R&D money, national-lab access, and no dilution whatsoever. There is no accelerator in this chapter that offers better economics to the founders who qualify. The constraint is purely eligibility: you need to be a scientist or engineer with a hard-tech venture, and the acceptance bar is high. If you are in the eligible population and you are considering a dilutive accelerator instead, apply here first. Note the FAQ caveat that life sciences and healthcare ventures should check fit before applying — the program skews toward energy, materials and physical-science hard tech.

Drawbacks. Narrow eligibility; two years is a long commitment; no investor-signal function comparable to a venture accelerator (though the National Lab affiliation carries weight with hard-tech investors and with SBIR reviewers).

NVIDIA Inception

Type: Corporate support program — no equity, no fee, no cohort Stage: Any, up to 10 years old Geography: Global Verification status: Fully verified on nvidia.com, September 2026.

Terms [Verified]. Free. No application fees, no deadlines, no cohorts, no equity taken. Benefits include preferred pricing on NVIDIA hardware and software, free self-paced technical training and discounted expert-led workshops, SDK and model-library access, free cloud credits from NVIDIA and partners, exposure to NVIDIA's VC network via "Inception Capital Connect" (eligibility-based), co-marketing, and curated networking events (NVIDIA, "Inception / Startups", accessed September 2026).

Eligibility [Verified]. Must employ at least one developer, have an official website, be officially incorporated, and be less than 10 years old. Excluded: consulting firms, cryptocurrency-associated companies, cloud service providers, resellers, and public companies.

[Analysis] Inception is not an accelerator and should not be compared to one. It is a developer-relations and channel program, and its value is almost entirely in hardware discounts and compute credits for companies whose cost structure is GPU-dominated. The honest framing: it is free, it takes ten minutes, and if you train or serve models you should be in it. The "VC network access" component is real but is not a substitute for an investor network — NVIDIA's incentive is to sell GPUs, and its introductions follow that incentive.

Plug and Play

Type: Corporate-innovation accelerator network Stage: Seed to Series B, varies by vertical Geography: Sunnyvale HQ plus dozens of global locations Verification status: Not verified against an official source at this date. Plug and Play's published program structure varies by vertical and corporate sponsor, and I could not confirm current terms.

The model. Plug and Play runs a very large number of vertical programs (fintech, mobility, insurtech, supply chain, health, energy and many others), each underwritten by corporate partners who pay to see startup dealflow. Most Plug and Play programs do not take equity and do not invest as a condition of participation, though Plug and Play Ventures invests separately in a subset.

[Analysis] The right mental model is "structured corporate business development," not "accelerator." The product sold to corporations is access to startups; the product sold to startups is access to corporations. If your bottleneck is getting into a procurement conversation with a large insurer, bank or automaker, this is a legitimate and low-cost channel. If your bottleneck is capital, product or investor signal, it is not. Scale is both the strength (enormous corporate reach) and the weakness (very large cohorts, limited attention, wide variance in how seriously corporate partners engage). Equity cost is usually zero, so the downside is mostly time. Verify terms for the specific vertical program.

The AI-model-provider credit programs (the biggest change of 2026)

Type: Vendor credit programs Verification status: [Reported], not verified on OpenAI's or Anthropic's own pages. This is important and I flag it strongly — see below.

What reportedly happened. In May 2026, according to reporting, Sam Altman announced at a closed-door Y Combinator event that OpenAI would offer YC companies $2 million in API credits, structured with uncapped SAFE equity agreements. Within days Anthropic countered with $500,000 per startup with no equity requirement — reportedly a 16x increase over a previous $30,000 allocation. OpenAI then revised its structure to $500,000 with no equity, plus an optional $1.5M tranche in exchange for equity. Across four YC batches of roughly 200 companies, the combined commitment could reach up to $800 million in credits per year (MLQ News, "OpenAI and Anthropic Pour Up to $800M a Year in Free Credits Into YC Startups").

Verification caveat. I could not confirm these specific figures on OpenAI's or Anthropic's own sites at this date. The secondary source is a news aggregator. Treat the numbers as directionally indicative of a real bidding war and not as confirmed terms. Confirm with the program directly, and note that these offers appear to be accelerator-gated — available to YC (and likely other top-tier accelerator) companies, not to the general public.

[Analysis] Three things about this matter more than the exact numbers.

First, it is the clearest evidence available that accelerator affiliation has acquired a new, concrete, non-signaling economic value in 2026. Historically the argument for YC's 7% was reputational — investor signal, network, brand. If a YC company can now access $500K–$2M of frontier-model credits that an unaffiliated company cannot, the equity is buying a deliverable input, not just a reputation. For an AI-native company with heavy inference costs, this could easily exceed the value of YC's own $500K.

Second, the equity-linked structure of the original OpenAI offer is a genuine warning sign. Credits granted in exchange for an uncapped SAFE are not free; they are a form of vendor financing that creates both dilution and technical lock-in. A founder who takes $2M of credits from a single model provider has made an architecture decision under financial duress. The right response is to build a provider-abstraction layer on day one regardless of whose credits you take, and to price the "free" credits at the cost of the switching you are foreclosing.

Third, this is a temporary phenomenon. Credit programs at this scale are marketing expenditure in a land-grab, and land-grabs end. Do not build a business whose unit economics only work on subsidized inference.

A3. Incubators and venture studios

The terminology in this space is used loosely and the distinctions are worth being precise about, because they determine how much of your company you keep.

How venture studio equity actually works

Venture studios ("company builders," "startup studios") originate ideas internally, validate them with in-house resources, then recruit a founding CEO to run the company. Because the studio contributes the idea, the initial capital, and often the first engineering team, studios take dramatically more equity than accelerators — commonly quoted ranges are 20–50% at formation, and the effective figure after the studio's fund also participates in the seed round is frequently at the high end or above.

The mechanics matter more than the headline. A typical structure:

1. Founding equity split. The studio holds a block of founder-equivalent common stock at incorporation, often 30–50%, in exchange for the pre-formation work. The recruited CEO and early team split the rest, subject to vesting.
2. The studio's fund then invests in the pre-seed/seed on investor terms, taking preferred stock on top of the founding block.
3. Subsequent rounds dilute everyone, but the studio's large starting position means it can maintain meaningful ownership through Series B without heroic follow-on.

[Analysis] The honest evaluation of a studio deal is: what would you have done with the 18 months and the idea if you had not taken it? If the answer is "not had the idea, not had the capital, not had a team," 30–40% is a legitimate price for something that otherwise would not exist. If the answer is "I have my own idea and I want funding," a studio is a strictly worse deal than any accelerator in this chapter, by a very large margin. Studios are an employment-adjacent path into entrepreneurship with unusual upside, not a funding option for people who already have companies.

The second honest point: a founder in a studio company is not fully a founder in the control sense. The studio typically holds board influence, sometimes board control at inception, and has strong opinions about the business it created. Founder-CEO turnover in studio companies is a real and underdiscussed phenomenon.

Active venture studios

Verification status for this list: [Reported / general knowledge]. Studio terms are almost never published; each entry below should be verified directly.

A note on studio performance claims. The studio industry publishes aggregate statistics claiming higher success rates than traditional startups — a commonly circulated figure holds that studio companies reach Series A at a much higher rate than the general population. [Analysis] These claims are produced by the Global Startup Studio Network and studio-affiliated researchers, which is a conflict of interest, and they compare a selected, well-capitalized, professionally-staffed population against all startups. The comparison is not apples to apples. Studio companies should reach Series A more often, because they start with money, a team, a validated idea, and a full-time operating partner. That says something real about the model's efficiency; it says nothing about whether you would do better inside a studio than outside one.

Corporate and university incubators

A large category that takes little or no equity: university incubators (see A5, usually free but affiliation-gated); corporate innovation labs, where the implicit cost is strategic entanglement — a corporate incubator that becomes your first customer also signals to that customer's competitors that you are captured, which is a real cost for horizontal tools; and civic/economic-development incubators offering below-market space, low cost and low signal, occasionally useful for physical-product companies.

A4. Startup competitions

Competitions are the only category in this chapter that offers non-dilutive capital with no application gatekeeping beyond merit. Their weakness is that the money is small relative to the time spent and the prestige rarely converts into investor interest outside the specific ecosystem.

Rice Business Plan Competition

Type: Intercollegiate startup competition Verification status: Verified on rbpc.rice.edu, September 2026.

The largest and richest student startup competition in the world. The 2026 competition ran April 9–11 and awarded more than $1 million in expected prizes; reporting on the actual 2026 event put total awards at $2.7 million (InnovationMap, 2026) — a gap worth noting, because the difference between the advertised pool and the actual total is made up of investment prizes and in-kind services, not cash. Prize structure for 2026 [Verified, RBPC Prizes]: 1st place $200,000 investment (Goose Capital), 2nd $100,000, 3rd $50,000, 4th–7th $5,000 cash each, plus a long list of category prizes (Aramco Innovator Prize $40,000; two Pearland EDC prizes of $25,000 cash; Amentum Space Technology $25,000 cash plus $25,000 in-kind). Every participating team receives at least $950. 42 graduate-led startups were selected to compete in 2026 (Rice News).

Critical eligibility note: RBPC is for graduate-student-led startups. It is not open to non-student founders. Note also that the headline prizes are investments, not grants — first place is $200,000 of investment from Goose Capital, which means equity.

Best fit. Graduate students with a company and a technical or scientific differentiator. The judging pool is genuinely investor-heavy, which makes RBPC one of the few competitions where the network is worth more than the money.

Other major competitions worth entering

Verification status for this list: [Unverified at this date] — I did not confirm 2026 prize amounts or eligibility for each. Check each official site.

Notable: the Chicago Booth New Venture Challenge has the strongest outcome record of any university competition, having launched Grubhub and Braintree among others — though again, selection effects.

[Analysis] The honest assessment of competitions. For most founders they are a poor use of time on an expected-value basis: the preparation cost is high, the win probability is low, and the prize money is small relative to a seed round. They are worth entering in three specific cases: (1) you are a student and the competition is the primary funding path available to you; (2) you are in a deep-tech or climate field where the judging panel contains the specific investors and corporates who matter in your sector, and the exposure is the prize; (3) the competition is a front door to something larger (MassChallenge's accelerator, Cleantech Open's mentor network). Entering competitions as a general fundraising strategy is a well-known failure mode — it produces founders who become good at pitching and not good at building.

A5. University entrepreneurship programs

University programs are the single most underused resource in the ecosystem, largely because founders assume they require enrollment. Many do; several important ones do not.

Verification status: [Partially verified]. Berkeley SkyDeck verified September 2026; others [Reported].

[Analysis] The three university-affiliated programs that matter most to non-students are Berkeley SkyDeck, Creative Destruction Lab, and Cornell Tech's Runway. CDL in particular deserves more attention than it gets: it is free, takes no equity, runs for nine months, and its mentor pool (successful founders and researchers, organized into objective-setting sessions every eight weeks) systematically culls companies that fail to hit targets. Roughly half the companies in a CDL stream are cut before the end. That culling is the product — it is one of very few programs that will tell you honestly that you should stop.

For everyone else, the practical advice is: university resources are typically gated by affiliation, and affiliation is often cheaper to obtain than people assume (executive-education enrollment, alumni status, a spouse's affiliation, a research collaboration). It is worth an email to the program director before assuming you are excluded.

A6. Government programs

Government programs are the largest source of non-dilutive capital and support available to startups, and they are systematically underused by venture-track founders — partly because the application burden is real, and partly because the venture ecosystem's culture treats grant-writing as unglamorous. The funding-instruments chapter covers the mechanics of these awards. This section covers the institutions.

SBIR and STTR (United States)

Type: Federal R&D grant/contract program — equity-free Verification status: Verified on sbir.gov and in legal reporting, September 2026. This section contains the single most important 2026 update in this chapter.

The program. SBIR (Small Business Innovation Research) and STTR (Small Business Technology Transfer) provide "equity free funding through federal agencies to American small businesses" (SBIR.gov, "About"). Eleven federal agencies participate, each running its own solicitations under SBA-coordinated rules. The program has run since 1982. STTR differs from SBIR in requiring a formal partnership with a research institution.

The 2026 lapse and reauthorization [Verified — critical]. SBIR/STTR authorization expired September 30, 2025 and lapsed from October 1 and remained lapsed for six months. Congress reauthorized the programs on April 13, 2026, extending them through September 30, 2031 (Crowell & Moring, "SBIR/STTR Programs Reauthorized After Six-Month Lapse"). Any guidance about SBIR written between October 2025 and April 2026 should be treated with suspicion, and any company that built a plan around SBIR during that window had a very bad six months. [Analysis] The lapse is itself the most important thing to understand about SBIR as a strategic resource: it is subject to congressional reauthorization risk on a recurring cycle, and it has now demonstrably failed to renew on time. Do not build a company whose survival depends on a Phase II award arriving on schedule.

Award sizes [Verified, as of April 2026]. Phase I up to $323,090; Phase II up to $2,153,927. Awards above these amounts require an SBA waiver (SBIR.gov). These caps are inflation-adjusted periodically, which is why the numbers are oddly specific.

What changed in the 2026 reauthorization [Verified]. Four things matter:

1. Strategic Breakthrough Awards — a new Phase II category for agencies spending over $100M annually on SBIR, with individual awards up to $30 million, either as a single award or milestone-triggered payments, over performance periods up to 48 months. This is a genuinely new scale of non-dilutive funding and is the most significant structural addition to SBIR in years.
2. Foreign risk screening — expanded vetting of applicants' foreign affiliations, investment connections, technology licensing and joint ventures with entities in "countries of concern," plus expanded watchlist exclusions. This has real consequences for startups with Chinese investors on the cap table or overseas R&D arrangements, and it is a reason to think about SBIR eligibility before taking certain investment.
3. Proposal limits from FY2027 — agencies will set their own caps on submissions per company, per solicitation or per topic, aimed at "SBIR mills" that submit enormous volumes of proposals.
4. Administrative standardization — standardized Phase I–III procedures and simplified contracts across agencies, with SBIR/STTR classification tracked in the Federal Procurement Data System.

Best fit. Deep-tech, defense, health, energy and materials companies whose R&D is genuinely pre-commercial. The money is free of dilution, and a Phase II award is a meaningful credential with hard-tech investors. Phase III — where the agency becomes a customer — is the actual prize and is the part founders most often ignore.

Drawbacks. Application burden is substantial (weeks of work for Phase I). Timelines are slow. Some agencies' programs are effectively captured by repeat winners. And the reauthorization risk is now demonstrated rather than theoretical. [Analysis] The classic failure mode is "grant treadmill" — a company that becomes good at winning SBIRs and never builds a commercial business. A useful discipline: if you cannot articulate who the paying customer is at the end of Phase II, the grant is a subsidy for a research project, not a startup.

SBA resources and SBDCs

Type: Free/low-cost counseling and training — no equity, no cost Verification status: Verified on sba.gov, September 2026.

The SBA's resource-partner network is the least glamorous and most broadly accessible support system in the United States (SBA, "Local Assistance"):

• Small Business Development Centers (SBDCs) — ~1,000 locations, usually hosted at universities; free business counseling, financial-projection help, market research, and — critically for founders — assistance with SBIR proposal preparation and government contracting registration.
• SCORE — volunteer mentoring by experienced (usually retired) business operators. Free.
• Women's Business Centers and Veterans Business Outreach Centers — targeted counseling and training.
• Regional Innovation Clusters, export assistance and US Export Assistance Centers for companies selling internationally.

[Analysis] The honest assessment: the mentor quality distribution is wide and skews toward traditional small-business experience rather than venture-scale technology. A SCORE mentor who ran a regional distribution business for thirty years is not the right advisor for a seed-stage AI company, and founders who expect otherwise come away dismissive of the whole network. But there are two things SBDCs do better than anyone: government-contracting navigation and SBIR proposal mechanics. Both are genuinely hard, genuinely learnable, and free at an SBDC. For a hard-tech founder planning an SBIR application, an SBDC visit is among the highest-return hours available.

State-level programs (United States)

Verification status: [Not verified individually]. There are fifty-plus distinct regimes and they change with each legislature.

The main categories:

• State SBIR matching programs — many states match federal Phase I awards (often $25K–$100K), which both increases the total and shortens the gap before Phase II.
• State venture funds and fund-of-funds — state-backed capital, frequently with in-state domicile or job-creation requirements.
• R&D tax credits — state-level credits on top of the federal credit; several states make theirs refundable for pre-revenue companies, which converts a tax credit into cash.
• Economic-development incentives — relocation grants, hiring credits, facility subsidies.

[Analysis] The single practical rule: state programs almost always require in-state domicile, in-state employment, or both, and these strings are more binding than founders expect. A grant that requires you to keep headcount in a state for five years is a constraint on where you can hire and where you can move. Read the clawback provisions. The cleanest state programs are the refundable R&D tax credits, which are money with almost no strings.

United Kingdom — Innovate UK

Verification status: [Partially verified]. Innovate UK's live opportunities were verified on ukri.org in September 2026; the individual program structures below are [Reported] and should be checked on the UKRI funding finder.

Innovate UK, part of UK Research and Innovation, is the UK's innovation agency and describes itself as backing "the nation's most promising deep tech businesses" (UKRI, "Innovate UK"). Its main instruments:

• Smart Grants — open-call, sector-agnostic grants for game-changing innovation, historically in the low hundreds of thousands to low millions of pounds, competitive rounds.
• Innovation Loans — patient debt for late-stage R&D, an unusual instrument with no clean US equivalent.
• Innovate UK Business Growth (formerly EEN/Growth Service) — free advisory support from innovation and growth specialists.
• ICURe — a commercialization program for university researchers testing whether their research has a market.
• Thematic competitions — live examples in September 2026 include Ultra Long Duration Energy Storage project development studies (up to £3M, deadline 30 September 2026) and Medicines Manufacturing Data Institute Phase 1 (up to £500,000, deadline 14 October 2026) [Verified on the UKRI page].

Complementary UK programs worth knowing: SEIS and EIS tax reliefs, which are the single biggest structural advantage of raising angel money in the UK — they make UK angel investment materially cheaper for investors than in most jurisdictions, and therefore make UK pre-seed rounds easier to fill. (Mechanics are covered in the funding-instruments chapter.) R&D tax credits for SMEs are also significant, though the scheme has been repeatedly tightened since 2023 in response to fraud.

European Union — EIC Accelerator

Type: Blended grant + equity Verification status: Verified on eic.ec.europa.eu, September 2026.

The EIC Accelerator is the EU's flagship instrument for single companies and the largest non-dilutive-plus-equity program in Europe (European Innovation Council, "EIC Accelerator").

• Grant: lump sum below €2.5 million for innovation activities completed within 24 months.
• Equity: €1–€10 million through blended finance (grant plus equity or convertible loan), delivered via the EIC Fund. Larger amounts are available through the STEP ScaleUp programme.
• Eligibility: single startups and SMEs including spin-outs; small mid-caps up to 499 employees for the investment component only; and — unusually — natural persons planning to establish an SME. Applicants must be from EU Member States or Horizon Europe associated countries. Third-country applicants may relocate before the full application. UK applicants can access the grant-only scheme.
• Target: Technology Readiness Levels 6–8, with market-creating or disruptive potential.
• Process: four steps — a short proposal (12-page form, 10-slide deck, 3-minute video; evaluated in 4–6 weeks), a full proposal (20 pages, submitted at a batching date), a jury interview for top-ranked proposals, then award and due diligence (2–6 months for the investment component).
• 2026 full-proposal cut-offs [Verified]: 7 January, 4 March, 6 May, 8 July, 2 September, 4 November, all at 17:00 Brussels time.

[Analysis] The EIC Accelerator is the best deal in Europe for capital-intensive deep tech, and it is also brutally slow and administratively heavy. The gap between short-proposal submission and money in the bank routinely exceeds a year once due diligence is counted. Success rates at the full-proposal stage have historically been in the single digits to low teens. The strategic use is as a parallel track to a private round, never as the primary plan — a company that runs out of runway waiting for EIC due diligence is a common story. The equity component also means the EIC Fund becomes a shareholder, and its governance and follow-on behavior are not like a private VC's.

Canada

Verification status: [Partially verified] on nrc.canada.ca, September 2026.

• NRC IRAP (Industrial Research Assistance Program) — "advice, connections, and funding" for Canadian SMEs, delivered through Industrial Technology Advisors who are assigned to companies. Named streams include financial support for technology innovation, funding to hire young graduates, clean technology, AI for SMEs, and defence/dual-use technologies (NRC, "Support for technology innovation"). IRAP does not publish standard amounts — funding is negotiated per project with an advisor, which is both its strength (flexibility) and its weakness (opacity).
• [Verified, important] The Government of Canada announced that the Canada Innovation Corporation will integrate NRC IRAP by 2026–27, with the program remaining at NRC until the transition. This is a live institutional change and founders should confirm the current administering body before planning around IRAP.
• SR&ED — Canada's R&D tax credit, and for many Canadian startups a larger source of cash than any grant. Refundable for Canadian-controlled private corporations, which makes it real money rather than a deferred benefit.
• Other: provincial programs (Ontario, Quebec and BC each run significant ones), MaRS Discovery District and Communitech as ecosystem hubs, and the Start-up Visa Program, which is unusual globally in offering permanent residence tied to designated-organization support.

Singapore

Verification status: [Not verified]. I was unable to retrieve current program terms from Startup SG or Enterprise Singapore at this date; the pages did not return substantive content. The programs below are named accurately but their 2026 amounts and eligibility rules are unverified — check startupsg.gov.sg directly.

• Startup SG Founder — a mentorship-plus-grant program delivered through Accredited Mentor Partners, historically structured as a government grant matched against founder capital.
• Startup SG Tech — proof-of-concept and proof-of-value grants for deep-tech commercialization.
• Startup SG Equity — government co-investment alongside qualified private investors, which is the mechanism that most changes Singapore's early-stage market: it multiplies private capital rather than replacing it.
• EntrePass — a work pass for foreign entrepreneurs, which is the reason Singapore is a genuinely viable base for non-resident founders in a way most jurisdictions are not.

[Analysis] Singapore's system is the most coherent national startup program in the world — the visa, the co-investment fund, the grants and the tax regime are designed as one system rather than as separate agency initiatives. The trade-off is a small domestic market and a heavy compliance culture. It works best for founders targeting Southeast Asia from a stable base, and for deep-tech companies that need government co-investment to make a round happen.

Israel — Israel Innovation Authority

Verification status: [Partially verified]. Program names and open status verified on innovationisrael.org.il, September 2026; amounts and royalty terms not published on the pages I could retrieve.

The Israel Innovation Authority (formerly the Office of the Chief Scientist) runs the world's most mature national innovation-funding apparatus. Programs relevant to early-stage founders, all listed as open year-round (Israel Innovation Authority, "Programs"):

• Ideation (Tnufa) — for very early entrepreneurs developing and validating a technological concept.
• Technological Innovation Incubators — franchise-like incubators that receive state backing to invest in and support early-stage companies.
• Venture Incubators Funding — venture creation and capital investment in technological startups.

[Analysis] Two structural features distinguish the Israeli model and are worth understanding even if you are not Israeli. First, grants are typically royalty-bearing: the state funds a large share of an approved R&D budget and recovers it through royalties on resulting revenue, meaning the money is non-dilutive but not free. Second, IP transfer restrictions apply — moving IP funded by IIA grants out of Israel triggers significant payments. This is a genuine constraint on acquisition structures and one that acquirers price in. Non-dilutive does not mean unencumbered, and the Israeli system is the clearest illustration of that principle anywhere.

A7. Cloud and compute credit programs

Verification status: AWS, Google Cloud, Microsoft and NVIDIA figures verified on official pages, September 2026. These programs change more often than any others in this chapter — quarterly is not unusual.

The three rules that matter more than the amounts:

1. The gap between the self-serve tier and the investor-linked tier is the largest single discount available to any startup. AWS: $5,000 versus $200,000. Google: $2,000 versus $350,000. The gate is an accelerator or investor relationship, not your technology. This is a concrete, quantifiable value of accelerator participation and it belongs in the equity calculation: roughly $200K–$350K of infrastructure, available only to affiliated companies.
2. Credits expire, usually in 12–24 months, and the card on file gets charged the day they do. Architect and forecast as though you were paying full price. The classic failure is a company whose gross margin is only positive on credits.
3. Do not multi-cloud to stack credits. The operational cost of running two clouds almost always exceeds the incremental credit value at seed stage. Pick one, take its largest available tier, and build a thin abstraction only where switching cost is genuinely asymmetric (model providers, yes; object storage, usually not worth it).

A8. Founder communities, angel networks and mentorship

Communities

On Deck's current state [Verified]. This is worth a specific note because On Deck's trajectory is widely misunderstood. On Deck raised large venture rounds in 2021 to build a broad portfolio of "On Deck X" fellowships (writers, angels, scale, etc.), then contracted sharply through 2022–23 with significant layoffs and program shutdowns. What survives, as of September 2026, is a focused product: ODF is "an intense one-week in-person experience" in San Francisco with 80–100 participants (over 50% technical), a lifetime alumni community of 3,000+, $800K+ in partner perks (Mercury, AWS, Stripe Atlas), and — importantly — it takes no equity. Self-reported: 1,000+ startups launched, $2B+ raised by alumni. ODF28 kicks off in Q1 2027 (ODF, accessed September 2026).

[Analysis] The compressed On Deck is a better product than the sprawling one was. A one-week, high-density, co-founder-finding event that takes no equity is a legible and honest offer. The thing to be clear-eyed about is that you are buying access to a room, and the value is entirely a function of who is in it — which varies by cohort and which you cannot assess in advance.

Founder Institute's equity structure [Verified], because it is unusual and often misrepresented. FI takes 2.5% via a warrant — reduced from 4% in 2022 — and founders do not sign the warrant at the start; they commit roughly two-thirds of the way through the program. The warrant activates only if the company raises above a threshold amount of outside capital, and FI gets no board seat or voting rights. The 2.5% is split: 0.5% to mentors, 1% to FI headquarters, 1% to local chapter leaders, with 60% of returns flowing back to local leaders and mentors; FI reports $8.5M+ distributed to date (Founder Institute, "Equity Collective").

[Analysis] FI is the most polarizing program in this chapter and the criticism is partly unfair and partly earned. Unfair: the fee is refundable before a defined checkpoint, the equity is a contingent warrant rather than stock, the terms are published openly (which almost no accelerator does), and the 200-city footprint reaches founders in cities with no other option. Earned: 2.5% plus a four-figure fee is expensive for a program with no investment, no meaningful investor signal, and highly variable local-chapter quality — the whole model depends on the local director, who is an equity-incentivized volunteer, not a professional investor. Best fit: a first-time founder in a city with no startup ecosystem who needs structure and a deadline more than capital or signal. Worst fit: anyone with a realistic shot at a funded accelerator, for whom FI is strictly dominated.

Angel networks

Angel groups are how organized angel capital reaches companies outside the top accelerators. Verification status: [Not verified individually] — membership rules, check sizes and activity levels vary and change.

• Angel Capital Association (ACA) — the US trade body; its directory is the reliable way to find legitimate groups in a region rather than relying on search results.
• Regional groups — Tech Coast Angels (Southern California), Band of Angels (Silicon Valley, the oldest), Hyde Park Angels (Chicago), Launchpad Venture Group (Boston), Golden Seeds (women-led companies), Pipeline Angels (women and non-binary founders), Keiretsu Forum (global network of chapters).
• Syndicate platforms — AngelList syndicates and similar structures let individual angels lead deals with pooled capital behind them. This is now a larger channel than traditional angel groups for software companies.

[Analysis] Two honest warnings about angel groups. First, process cost. Many groups run a multi-month screening, pitch, and due-diligence gauntlet for what may end up being a $150K aggregate check split across twenty members with twenty sets of paperwork — a terrible ratio of founder time to capital. Ask up front: what is the median check size the group has actually written in the last year, and how long from first meeting to wire? Second, any group that charges founders to pitch is a red flag. Legitimate angel groups charge their members dues; they do not charge founders application or presentation fees. This is one of the few bright-line tests in the whole ecosystem.

Coworking and the physical ecosystem

The physical layer matters less than it did in 2019 and more than the remote-work consensus of 2021 suggested.

• WeWork emerged from Chapter 11 bankruptcy in 2024 as a smaller, privately held company [Reported]. The lesson: do not sign a long lease with a coworking operator whose own balance sheet is a going-concern question. Month-to-month is correct at seed stage regardless of the annual-commitment discount. Industrious, Regus/IWG and regional operators occupy the general-purpose market.
• Startup-specific spaces — 1871 (Chicago), Cambridge Innovation Center, Station F (Paris, the largest startup campus in the world), university-affiliated spaces — differ from generic coworking in that the other tenants are the product.
• Sector-specific infrastructure is the category that actually matters: wet-lab space (LabCentral, BioLabs, JLABS), hardware prototyping shops, GPU clusters. For a biology or hardware founder, shared lab or fab access is a first-order determinant of capital efficiency — the reason SOSV and Activate have real, non-reputational value.

[Analysis] The honest position in 2026: for a fully remote software team coworking is a cost with modest return; for an in-person team it is a commodity; for a wet-lab or hardware company it is existential. The only universally valuable physical asset is proximity to the specific people you need, which is a question about the city, not the building.

PART B — DO ACCELERATORS ACTUALLY WORK?

B1. What the academic evidence can and cannot establish

The central empirical problem is straightforward and has never been fully solved: accelerators select the companies they admit, and the companies they admit are systematically different from those they reject. Any comparison of accelerator graduates to non-participants measures selection plus treatment, and the two are extremely hard to separate.

What the literature has established with reasonable confidence:

1. Top-tier accelerators are associated with better outcomes; the effect shrinks dramatically outside the top tier. The most-cited body of work here is by Susan Cohen, Yael Hochberg, and collaborators, whose research on accelerator cohorts consistently finds a wide quality dispersion — a small number of programs are associated with meaningfully better funding and survival outcomes, and the median program shows weak or no measurable effect. Hochberg's "Accelerating Entrepreneurs and Ecosystems: The Seed Accelerator Model" (in Innovation Policy and the Economy) is the standard reference for the model's mechanics and for the observation that the accelerator population is extremely heterogeneous.

2. The strongest identified mechanism is speed, not success. Research by Benjamin Hallen, Christopher Bingham and Susan Cohen ("Do Accelerators Work? If So, How?", Organization Science) found that accelerator participation was associated with companies reaching milestones faster — raising money sooner, or failing sooner — rather than simply succeeding more often. [Analysis] This is the most useful and most underrated finding in the entire literature, and it reframes the value proposition correctly. An accelerator compresses the time to learn whether your company works. For a founder, faster failure is genuinely valuable — it returns your most finite asset, your years. But "accelerates outcomes" is a different claim from "improves outcomes," and programs market the latter while the evidence better supports the former.

3. Regression-discontinuity studies around admission thresholds find positive but modest effects. Work by Juanita González-Uribe and colleagues on accelerator and entrepreneurship-program admissions, using applicants scored just above and just below a cutoff, has found effects on subsequent fundraising and survival that are real but far smaller than the raw comparison of participants to non-participants implies. [Analysis] The gap between the raw comparison and the discontinuity estimate is, roughly, the size of the selection effect. It is large.

4. Accelerators change ecosystems, not just companies. There is reasonable evidence that an accelerator's entry into a regional ecosystem increases local seed-stage financing activity generally, including for non-participants. This is the strongest argument for public subsidy of accelerators and the weakest argument for any individual founder joining one.

What the literature has not established: whether YC specifically causes its outcomes (no credible study isolates its treatment effect, and its admissions are not a scoreable threshold); the counterfactual for any individual company, since effects are almost certainly heterogeneous — negative for some, large for others; and long-run effects, since most studies observe only 2–5 years of fundraising and survival.

[Analysis] The honest synthesis. Accelerators produce a real but modest average treatment effect, concentrated in the top programs, operating primarily through (a) speed, (b) investor access, and (c) forced accountability. The enormous outcome differences between accelerator graduates and everyone else are mostly selection. Anyone who tells you that YC "makes" companies successful is wrong, and anyone who tells you YC does nothing is also wrong. The truth is that YC is very good at identifying founders who were going to do well, and then modestly improving their odds and substantially compressing their timeline.

B2. The quality spread and the "accelerator for accelerator's sake" problem

There are, by most counts, several thousand entities worldwide describing themselves as accelerators. The number that provide value in excess of what they charge is, plausibly, a few dozen.

The failure mode has a specific and recognizable shape. A regional economic-development agency, a corporation, or a university decides an accelerator is the answer to a problem — jobs, innovation pipeline, brand relevance — and creates one. It has: a three-month program, a demo day, a mentor list, a small check, and a name. What it usually does not have is an investor audience that cares about its demo day, mentors with directly relevant operating experience, or deal flow good enough that the cohort is a valuable peer group. These three things are the entire product. Without them, the program is a co-working space with a curriculum.

Diagnostic questions that separate real programs from theater — ask them directly, and treat refusal to answer as an answer:

1. Of your last three cohorts, what fraction raised a priced round within 18 months of demo day? (Not "raised money" — priced round. SAFEs from friends do not count.)
2. Name five investors who attended your last demo day and have written a check to one of your companies.
3. How many of your mentors have operated a company at the stage and in the sector of the companies you admit?
4. What is your program's survival rate at three years?
5. Who funds the program, and what do they get? (A corporate sponsor buying dealflow, a government buying job statistics, and an investor buying equity all produce different incentives.)
6. What happens to companies that are clearly failing mid-program? Do you tell them?

[Analysis] The strongest signal of a good program is that it will say negative things about its own companies and about itself. CDL publicly culls half its cohort. HF0 says it only backs ten teams. Programs that only produce superlatives are marketing organizations.

The cost of a mediocre accelerator is higher than founders assume, because it is not just equity: three months of calendar time at the stage when time is most valuable, a cap-table entry every future investor will ask about, and a demo day that produces no term sheets, which reads to the market as a failed process. A weak accelerator can leave you less fundable than you were before you joined.

B3. Bootstrapper versus venture-track: two entirely different calculations

Almost all accelerator advice implicitly assumes venture ambition. For a bootstrapper the calculation inverts.

For a venture-track founder, the accelerator's core product is access to the capital market. You are buying a warm, credible, time-compressed introduction to the investors who will fund your next three rounds. Equity is the correct currency for this because the thing you are buying is denominated in future equity value. If you already have that access — you are a repeat founder, you worked at a company whose investors know you, you have a network — the value falls sharply and the 7% becomes hard to justify. The people who benefit most from YC are precisely the people who have no other way in: first-time founders, international founders, technical founders with no commercial network, founders outside the coastal ecosystems.

For a bootstrapper, the same product is nearly worthless. You are not raising a round, so investor access is irrelevant, and the accelerator's incentive structure pushes the opposite way — toward growth metrics that justify a round, hiring ahead of revenue, and a demo-day narrative. Giving 7% (or 2.5% in a warrant, or anything) from a company you intend to own for a decade is a permanent cost against a benefit you will not use.

What bootstrappers should use instead: MicroConf and Indie Hackers (peer communities calibrated to getting from $0 to $1M ARR profitably); SBDCs and SCORE, whose traditional-small-business orientation is a feature here rather than the bug it is for a venture founder; non-dilutive government money (SBIR if you are technical, R&D tax credits everywhere); cloud credits at the self-serve tier ($2K–$5K, no investor relationship required); and competitions, the only structured category offering cash without equity.

[Analysis] The one case where a bootstrapper should consider an accelerator is when the program takes no equity and offers something specific and physical — Activate's $300K and lab access, an SBIR-focused program, a corporate accelerator that delivers a first enterprise customer. "Community and mentorship" is not a reason for a bootstrapper to give up equity, because both are available free.

B4. How the value proposition shifted in 2025–26

Four changes have reshaped what an accelerator is worth, and they do not all point the same direction.

1. Capital concentrated violently in AI, which inverted who needs the signal. With the large majority of venture dollars flowing to AI companies and deal counts falling even as dollar volumes hit records, the seed market bifurcated. For a non-AI company, a YC signal now does more work than it used to, because ambient investor willingness to look at your sector is lower. For an AI company it is less differentiating, because the batch is full of AI companies and so is everyone else's.

2. The cost of building collapsed, which shortened the useful length of a program. When a competent team can ship a working product in weeks, the value of three months of structured "build" time falls. What remains valuable is distribution, customers, and capital — the things AI has not made cheap. [Analysis] This is the strongest argument for the short-and-intense formats (Neo's three months plus bootcamp, SPC's eight weeks, ODF's one week) and against six-month programs for software companies. Alchemist's six months makes sense for enterprise sales cycles; six months makes little sense for a consumer app.

3. Compute credits became a concrete, accelerator-gated asset. The reported OpenAI/Anthropic offers to YC companies, and the $200K–$350K cloud tiers gated behind investor relationships, mean accelerator affiliation now carries a quantifiable non-reputational benefit for the first time. For an inference-heavy AI company, this may be the single largest component of the accelerator's value, and it did not exist in 2023. It is also the component most likely to disappear, because it is a function of a subsidy war.

4. Competition on terms intensified. Neo's uncapped $750K, SPC's $400K/7% plus $600K guaranteed, Techstars' move to $220K with only 5% fixed, and a16z Speedrun's $1M all reflect a market where the best founders have options. [Analysis] The competitive pressure is real and it is good for founders, but it has a second-order effect worth naming: it compresses accelerator economics, which means programs must be even more selective to make their fund math work, which means selection effects get stronger, not weaker. A program writing $750K checks into 30 companies a year needs outcomes that only the top of the applicant pool can deliver. The terms get better and the door gets narrower at the same time.

B5. The bottom line

Worth the equity, for the right founder:

• Y Combinator — if you are venture-track, lack investor access, and can relocate. The signal is real and remains the best-priced version of that product.
• South Park Commons — the most founder-favorable headline terms among pre-idea programs, with a genuinely non-transactional community.
• A strong, specific Techstars program — where the Managing Director and the corporate partner match your sector.
• SOSV (HAX / SOSV NY / SOSV SF) — because lab and prototyping access is a physical asset you cannot get elsewhere.
• Alchemist — if your bottleneck is a first enterprise customer.

Better than the equity-taking options, if you qualify — take these first:

• Activate — $300K+ over two years, zero equity, national-lab access. There is nothing better for eligible hard-tech scientist-founders.
• Creative Destruction Lab — nine months, no equity, honest culling.
• SBIR/STTR — reauthorized through 2031, now with $30M Strategic Breakthrough Awards.
• NVIDIA Inception — free, ten minutes, real hardware discounts.
• EIC Accelerator — for European deep tech, if you can survive the timeline.

Probably not worth it:

• 500 Global's Flagship at 6% for $150K, for a US founder with US investor access. The terms have not kept pace with the brand's decline.
• a16z Speedrun at a reported 10%, unless you specifically want a16z as your lead — and you should price the signaling risk if they pass.
• Founder Institute, for anyone with a realistic shot at a funded program.
• Any regional accelerator that cannot answer the six diagnostic questions in B2.
• Any accelerator at all, for a founder who intends to bootstrap.

And the thing to keep clearly in mind throughout: every impressive statistic in Part A is a statement about who a program admits. HF0's $82M average demo-day valuation, YC's $600B portfolio, Alchemist's $5B raised — all of these are measurements of selection, computed over survivors, published by interested parties. The question is never "is this program good?" It is "what does this program do for a company like mine that I cannot do myself, and is that worth more than the equity?" For most founders and most programs, the honest answer is no. For a specific minority, it is emphatically yes.

Sources

All URLs accessed on or before September 15, 2026.

Y Combinator

• Y Combinator — The YC Deal
• Y Combinator — Apply
• Y Combinator — YC's $500,000 Standard Deal
• Y Combinator — A New Standard Deal
• Ellenox — Y Combinator Statistics and Insights (February 2026)
• ValueAdd VC — YC Acceptance Rate 2026

Techstars

• Techstars — Investment Terms Update (April 17, 2025)
• Techstars — 2024 Investment Terms Announcement
• Techstars — Accelerators
• Colorado Sun — Techstars moving HQ out of Colorado and ending its Boulder accelerator (February 2024)
• Colorado Sun — Techstars Colorado brings entrepreneur accelerator back to Boulder (October 2024)
• GeekWire — Techstars Seattle is shutting down (2024)
• Technical.ly — Techstars will end JPMorgan-backed Washington accelerator
• TechCrunch — Techstars CEO defends changes (February 2024)

Other accelerators and residencies

• 500 Global — Flagship Accelerator
• Antler — Antler in the US
• Antler — Antler Singapore improved terms and 6-week residency
• Built In Austin — Antler updates its US investment terms
• Entrepreneur First — FAQs
• Entrepreneur First — London FAQs
• Alchemist Accelerator — Programs
• HF0 — Facts
• HF0 — Home
• South Park Commons — SPC Founder Fellowship, Fall 2026
• South Park Commons — SPC Founder Fellowship Spring 2026
• TechCrunch — Ali Partovi's Neo looks to upend the accelerator model with low-dilution terms (February 19, 2026)
• a16z — Speedrun

Sector-specific and deep tech

• SOSV — Application FAQs
• SOSV — About
• IndieBio / SOSV
• HAX — FAQ
• Activate — The Fellowship
• Activate — Activate Welcomes 50 New Fellows to Cohort 2026
• Activate — FAQ
• NVIDIA — Inception / Startups

Venture studios

• High Alpha — What is a Venture Studio?
• Alloy Partners — Venture Studio vs. Venture Capital (2026)

Competitions and universities

• Rice Business Plan Competition — Home
• Rice Business Plan Competition — Prizes
• Rice News — 42 graduate-led startups selected to compete at RBPC (2026)
• InnovationMap — Rice Business Plan Competition awards $2.7M to 2026 student teams
• Berkeley SkyDeck

Government programs

• SBIR.gov — About
• Crowell & Moring — SBIR/STTR Programs Reauthorized After Six-Month Lapse
• CSIS — SBIR and STTR Reauthorization and the Future of Small Business Innovation
• IEDC — Congress Reauthorizes SBIR and STTR (April 2026)
• SBA — Local Assistance
• UKRI — Innovate UK
• European Innovation Council — EIC Accelerator
• NRC Canada — Support for technology innovation (IRAP)
• Startup SG
• Israel Innovation Authority — Programs
• Israel Innovation Authority — Home

Cloud and AI credits

• AWS Activate — Credits
• Google Cloud for Startups
• Microsoft for Startups
• MLQ News — OpenAI and Anthropic Pour Up to $800M a Year in Free Credits Into YC Startups

Communities and networks

• ODF (On Deck Founders)
• Founder Institute — Equity Collective
• Founder Institute