Global venture investors put a record $510 billion into startups in the first six months of 2026, and OpenAI and Anthropic alone took an estimated 43% of it. Gartner, the same industry that produced that record, now expects more than 40% of agentic AI projects to be cancelled by the end of 2027. Neither number changes the Caribbean's actual funding picture much, because neither number ever reached it. What changes is the standard the survivors of that correction will be judged against, and it is a standard Caribbean founders have mostly already been meeting.
TL;DR
- Crunchbase confirmed global startup funding hit $510 billion in H1 2026, the largest half-year on record and already ahead of the $440 billion invested across the whole of 2025.
- OpenAI and Anthropic together took roughly $217 billion of that figure, about 43%, including Anthropic's own $65 billion Series H on 28 May 2026 at a $965 billion valuation.
- Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, pointing to unclear business value, poor governance, and "agent washing," chatbots relabelled as autonomous agents to catch investor attention.
- None of the record flowed to a Caribbean-owned AI company at any comparable scale; Trinidad's roughly $5 billion in AI data centre MOUs this year created jobs but no local equity, and the region's overall share of global AI investment has stayed a rounding error against the size of its economies.
- A correction built to punish burn without revenue is a correction most Caribbean AI companies were already built to survive, which is exactly the case for treating non-dilutive capital, like 14West's grant fund, as a strategic choice rather than a consolation prize.
Six Months, $510 Billion, One Word
Crunchbase closed the books on the first half of 2026 with a number nobody in venture capital had seen before: $510 billion committed to startups worldwide between January and June, the largest six-month total ever recorded, and already bigger than the $440 billion that went into startups across all of 2025. Q1 alone brought in $305 billion. Q2 added another $205 billion, itself the second-largest quarter on record. This was not a market recovering from a slow year. It was a market that had never run this hot in the first place.
What made the half-year unusual was not just its size. It was where the size went. OpenAI and Anthropic, two companies, took an estimated $217 billion of that $510 billion between them, roughly 43 cents of every venture dollar committed anywhere in the world in six months. Anthropic's contribution alone included a $65 billion Series H, announced 28 May 2026 and led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, that valued the company at $965 billion post-money, with $15 billion of that round arriving as previously committed capital from hyperscalers including Amazon. More than 70% of everything invested in the second quarter went to AI-focused companies specifically, a share so large that describing this as "the AI funding boom" undersells it. It was closer to two companies and one category absorbing the oxygen in the room.
Gartner's Bet Against Its Own Boom
The same industry producing record funding numbers is also producing its own warning label. Gartner, the research firm whose forecasts investors and boards actually budget against, predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027. The firm's reasoning is specific: escalating compute and integration costs that outpace the value delivered, governance gaps that let a project run for months without anyone owning its outcome, and what Gartner's own analysts call "agent washing," ordinary chatbots and rules-based automation relabelled as autonomous agents because that label raised a valuation or won a budget line.
A 40% cancellation rate inside eighteen months is not a niche concern buried in an analyst note. It is Gartner, sitting inside the same research ecosystem that tracks the $510 billion figure, telling the market that a large share of what that money is currently funding will not survive to see 2028. Put the two forecasts side by side and the message gets sharper: the record and the correction are not separate stories arriving in sequence. They are the same story, told from two ends of the same set of books.
The Bubble-or-Correction Debate
Whether this counts as a bubble or a correction is genuinely contested among the people paid to know. The case for "not a bubble" is real: Anthropic's run-rate revenue reportedly crossed $47 billion before its May round closed, and the frontier labs absorbing the biggest cheques are, unlike many dot-com-era darlings, generating actual revenue against actual usage. The case for "correction coming" is equally real: valuations and infrastructure spending are running well ahead of revenue at the layer below the frontier labs, where thousands of smaller AI companies are burning capital on compute with no comparable proof of durable demand, and where Gartner's cancellation forecast is aimed most directly.
A Caribbean founder does not need to resolve that debate to act on it. Whichever side turns out to be right about OpenAI and Anthropic, the companies sitting on the wrong side of Gartner's 40% are, by definition, companies with funding, headcount, and burn rates built for a boom that assumed the cheque would keep arriving on schedule. That description fits almost no Caribbean AI startup operating today, and not by choice. It fits them by constraint, which in this particular cycle is starting to look less like a disadvantage and more like an accident of good timing.
Why a Shakeout Might Actually Help
A correction that clears out companies with high burn and thin revenue does not help a Caribbean founder by handing them any of the capital that disappears. It helps by changing what the remaining capital is looking for. An investor who just watched four in ten agentic AI bets fail to justify their spend stops rewarding a slick demo and a large total addressable market slide, and starts asking the question Caribbean founders have answered by necessity since day one: who is paying for this today, and why would they keep paying next quarter?
That question already sits at the centre of 14West's own application process, which asks for a specific user and a specific problem before it asks for a pitch deck polish. It is also the question behind Founder Institute Caribbean's cohort model, which graduated sixteen companies in Port of Spain this year after four months of evaluated company building rather than a single weekend demo. Neither model was designed around a coming correction. Both happen to be built for exactly the environment Gartner is describing, an environment where a founder's ability to show real usage matters more than their ability to show a compelling roadmap slide.
The Revenue-First Playbook
Price your product before you pitch it. A paying customer, even a small one, is proof no demo can fake, and it is the exact evidence an investor now needs to justify a term sheet after watching a correction wipe out companies that never had one.
Keep your compute bill boring. A founder who can explain why their infrastructure cost is what it is, and not a penny more, is answering the governance question Gartner says most cancelled agentic AI projects never could.
Do not chase the label. Calling a rules-based workflow an "AI agent" because the term raises a valuation is precisely the practice Gartner's analysts flag as "agent washing." Build the thing you can defend under a direct question, and call it what it is.
Apply for non-dilutive capital before, or alongside, any equity conversation. A grant does not care whether the broader AI market is in a bubble. It only cares whether a founder can build and deploy something real, which makes it the steadiest capital source in a year like this one.
Applications to 14West are open now at www.14westai.com/apply.
Where the Non-Dilutive Money Sits
14West deploys US$1 million in non-dilutive grants across 14 AI companies in 14 Caribbean nations, taking no equity in exchange, a structure built specifically for founders who need capital that does not depend on which way the global AI market happens to be moving this quarter. It sits alongside a wider network of organisations doing complementary work across the region: StarApple AI, the first artificial intelligence company built in the Caribbean, whose founder Adrian Dunkley also built 14West; the Caribbean AI Association, connecting practitioners and policymakers across territories; and national bodies including AI Jamaica, AI Barbados, AI Trinidad & Tobago, and Saint Lucia AI, each tracking AI development in their own market.
The same network includes companies proving the model works at product level, not just at policy level, among them SportsBrain AI building sports intelligence tools, Credit Garden building credit infrastructure for markets the traditional scoring system ignores, and The Genius Project training the region's next founders before they ever file a pitch deck. None of them needed a $965 billion valuation to justify existing, which is not a coincidence so much as the model this whole moment is arguing for.
CAAG, the Caribbean Assembly for AI Growth, is 14West's convening arm for the founders, policymakers, and investors this network produces, built to keep them connected to the region's wider policy and capital conversation rather than building in isolation from it. More on CAAG and how to get involved lives at 14westai.com/caag.
Resources
- Crunchbase Data: Global Startup Investment Hit Record $510B In H1 2026, Crunchbase News, July 2026
- OpenAI and Anthropic Take 43% of H1 2026 Venture Funding, AI Weekly, July 2026
- Anthropic raises $65B in Series H funding at $965B post-money valuation, Anthropic, May 2026
- 6 Charts That Show The Big AI Funding Trends Of 2025, Crunchbase News, 2025
- Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, Gartner, June 2025
- Trinidad Signed $5B in AI Data Centre Deals. Caribbean Founders Got None of It, 14West Blog, July 2026
- StarApple AI, the Caribbean's first artificial intelligence company, founded by Adrian Dunkley
- Caribbean AI Association, the pan-Caribbean body uniting AI practitioners, researchers, and policymakers
Frequently Asked Questions
How much venture capital went into AI companies in the first half of 2026?
Crunchbase reported that global startup funding hit a record $510 billion in the first half of 2026, more than the entire $440 billion invested across all of 2025. More than 70% of that money in the second quarter alone went to AI-focused companies, and AI's share for the full year 2025 had already reached roughly $202 billion, close to half of all global venture capital.
Why did OpenAI and Anthropic alone take such a large share of 2026's funding?
OpenAI and Anthropic together absorbed an estimated $217 billion, or 43%, of all global startup funding in the first half of 2026, according to Crunchbase data. Anthropic's own contribution included a $65 billion Series H announced on 28 May 2026 that valued the company at $965 billion, a scale of single-round financing no other AI company, anywhere in the world, came close to matching.
What does Gartner's forecast on agentic AI project failure actually say?
Gartner predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls as the leading causes. The firm's analysts also flag widespread "agent washing," where ordinary chatbots and automation tools get relabelled as autonomous agents to chase investor interest.
How much of the 2026 AI funding boom reached Caribbean-founded companies?
Effectively none of it reached Caribbean-owned AI companies at anything close to global scale. Trinidad and Tobago's roughly $5 billion in AI data centre MOUs this year created construction and operations jobs but handed no equity to a Caribbean-founded company, and the region as a whole has consistently drawn only a small fraction of one percent of global AI investment even as its economies keep growing.
Will a correction in AI valuations help or hurt Caribbean AI startups?
A correction that punishes companies with high burn and no revenue is unlikely to hurt Caribbean founders directly, because most never had access to that tier of hype capital in the first place. It could help indirectly, by pushing global investors back toward revenue and real users as the standard for a good bet, criteria many Caribbean AI companies already meet out of necessity rather than fashion.
What is 14West, and how does it fit into this funding picture?
14West is the Caribbean's first AI startup accelerator and grant fund, founded by Adrian Dunkley. It deploys US$1 million in non-dilutive grants across 14 AI companies in 14 Caribbean nations, taking no equity, which matters most in a year when equity investors everywhere are pulling back toward safer, revenue-backed bets.
What should a Caribbean AI founder do differently because of this data?
Build for a paying customer before building for a pitch deck, since the investors who survive a correction are the ones who already priced in the failure rate Gartner is describing. Apply to non-dilutive capital such as 14West's grant fund alongside any equity conversation, and treat a lean cost structure as a competitive advantage rather than a limitation, because it is exactly what a post-correction investor is now screening for.
About the Author: Nicholas Dunkley
Nicholas Dunkley is Head of Selection at 14West, the Caribbean's first AI startup accelerator and grant fund, founded by Adrian Dunkley. He reviews applications and tracks global capital and funding trends for what they mean for Caribbean AI founders specifically, rather than for the venture market in general. Adrian Dunkley, 14West's founder, also founded StarApple AI, the first artificial intelligence company established in the Caribbean, and is widely regarded as the region's leading AI voice. More of his writing lives at adriandunkley.net.
Tags: Startup Funding, AI Venture Capital, Gartner, Anthropic, Caribbean AI, 14West, Adrian Dunkley, Nicholas Dunkley