AI Coding Funding: A Quarter-by-Quarter Analysis

All funding deals in the AI code assistant market over the last years

Below is the table listing all the deals. You can find our methodology at the end of this page.

If you want a deeper understanding of the market and its current dynamics, get our report covering the AI Coding Market.

Is AI code assistant funding actually growing right now?

AI code assistant funding is extremely uneven right now, with short bursts of huge scale-up capital separated by much smaller quarters.

Across the five quarters we studied, funding moved from $928.85 million in Q2 2025 to $32.09 million in Q3, jumped to $2.383 billion in Q4, then fell to $130 million in Q1 2026 and $1.75 million in Q2. Looking at that line alone makes the market almost impossible to read.

Once we remove rounds above $50 million, the pattern gets clearer. Underlying funding increased from $28.85 million in Q2 2025 to $32.09 million in Q3 and $82.7 million in Q4. That was a real improvement during 2025. It then dropped to $5 million in Q1 2026 and $1.75 million in Q2.

Today, the market can still support enormous financings when a company reaches serious scale. The broader flow of deals is much thinner and far less consistent.

Quarter	Deals	Total funding	Funding excluding >$50M	Median round
Q2 2025	4	$928.85M	$28.85M	$14.00M
Q3 2025	3	$32.09M	$32.09M	$2.59M
Q4 2025	5	$2.383B	$82.70M	$32.00M
Q1 2026	2	$130.00M	$5.00M	$65.00M
Q2 2026	1	$1.75M	$1.75M	$1.75M
Full period	15	≈$3.48B	≈$150.4M	-

Was Q2 2025 basically just Cursor?

Cursor explains almost the entire Q2 2025 funding boom in AI code assistants.

The quarter produced $928.85 million across four financings, and Cursor alone raised $900 million in its Series C. The other three companies combined raised $28.85 million: Unblocked brought in $20 million, Tabnine $8 million and Nia $850,000.

Cursor was already operating on a completely different scale. When the company announced the round, it said annual recurring revenue had passed $500 million and more than half of the Fortune 500 were using the product. A financing close to $1 billion makes far more sense against that commercial trajectory than as a benchmark for a typical AI coding startup.

The useful takeaway from Q2 is the gap inside the market. Several companies could raise money, but almost all the capital went to one business that had already reached hundreds of millions of dollars in recurring revenue.

Did AI code assistant funding really collapse in Q3 2025?

AI code assistant funding became much smaller in Q3 2025, but the underlying market held up far better than the 96.5% headline drop suggests.

Cursor's previous $900 million round disappeared from the comparison, leaving three Q3 deals worth $32.09 million. Excluding large rounds from Q2, the comparable funding base was $28.85 million, so underlying capital actually increased by about 11%.

The mix changed much more than the amount. Cline raised $27 million, CodeKarma $2.5 million and Lightbase approximately $2.59 million. Two of those three financings were first financings, and every deal was Seed or Series A. The median round consequently fell from $14 million to $2.59 million.

Q3 was a quarter of younger companies and smaller checks. Investors were still funding AI coding startups; they just were not funding another Cursor-sized company that quarter.

Was Q4 2025 a real AI coding funding recovery or another Cursor spike?

Q4 2025 was the strongest broad quarter we found, even before Cursor's $2.3 billion Series D entered the picture.

Five companies raised money, up from three in Q3. The four non-Cursor rounds totaled $82.7 million, more than two and a half times all Q3 funding. Code Metal raised $36.5 million, CoreStory $32 million, Kilo Code $8 million and Nozomio $6.2 million.

The median round also climbed from $2.59 million to $32 million, which tells us several companies were getting larger checks at the same time. Cursor then pushed the headline total to $2.383 billion.

Cursor's own business had continued to move unusually fast. By the Series D announcement, the company said annualized revenue had crossed $1 billion, millions of developers were using Cursor and the team had grown beyond 300 people.

Q4 was a genuine improvement across the market, then Cursor amplified it enormously. The disappointment came afterward: 2026 failed to carry that breadth forward.

Q4 2025 financing	Category	Round size	Stage
Cursor	IDE Code Assistants	$2.300B	Series D+
Code Metal	Refactoring AI Tools	$36.5M	Series A
CoreStory	Code Understanding Tools	$32.0M	Series A
Kilo Code	Developer Coding Copilots	$8.0M	Seed
Nozomio	Code Understanding Tools	$6.2M	Seed
Q4 total	-	$2.383B	-
Q4 excluding Cursor	-	$82.7M	-

Are investors writing bigger AI coding checks now?

AI coding checks briefly became much bigger in late 2025, but that increase has not held across subsequent quarters.

The median round fell to $2.59 million in Q3 2025 and then jumped to $32 million in Q4. That Q4 increase was meaningful because several companies contributed to it: Code Metal and CoreStory both raised more than $30 million.

Q1 2026 looks even stronger at first glance, with a $65 million median. There were only two deals, however, so that figure simply sits halfway between Code Metal's large round and Adronite's $5 million Series A. It tells us very little about what an ordinary company could raise.

By Q2 2026, the only financing was GitHits at $1.75 million. Q4 still looks like a short period when the middle of the market genuinely moved up in size, not the start of a steady rise in AI coding round sizes.

Are investors giving up on new AI coding startups?

New AI coding startups are still getting funded, although the really large checks are going overwhelmingly to companies that have already raised before.

First financings represented 25% of deals in Q2 2025, roughly 67% in Q3, 20% in Q4, zero in Q1 2026 and 100% in Q2. Those percentages jump around, but new companies clearly kept entering the market.

We saw Nia raise $850,000, CodeKarma and Lightbase appear in Q3, Kilo Code raise its first $8 million in Q4, and GitHits close a $1.75 million round in Q2 2026.

The size gap is huge. Across all five quarters, first financings contributed only about $15.7 million of roughly $3.48 billion raised. Young startups are still getting a shot; reaching the point where investors will commit $50 million or $100 million is much rarer.

That is why the market can look early-stage by deal count and late-stage by dollars at exactly the same time.

Is code understanding becoming the biggest AI coding startup theme?

Code Understanding Tools have become the most consistent source of new AI coding deals in our data, even though they still attract relatively modest amounts of capital.

The category appears in every quarter. Nia and Unblocked were funded in Q2 2025, CodeKarma and Lightbase in Q3, Nozomio and CoreStory in Q4, Adronite in Q1 2026 and GitHits in Q2.

There is a good product reason for that persistence. AI-generated code is improving quickly, but developers still struggle with context: understanding existing repositories, dependencies, documentation, architecture and why old code behaves the way it does.

Stack Overflow's 2025 Developer Survey captures that gap quite well. It found that 84% of respondents were using or planning to use AI tools in software development, while 46% distrusted the accuracy of AI output and 66% said one of their biggest frustrations was getting AI-generated solutions that were almost right.

That leaves room for companies trying to make coding models better informed before they generate or modify anything. Investors keep backing that idea, although most of the rounds remain small.

Quarter	Code Understanding deals	Companies	Capital
Q2 2025	2	Nia, Unblocked	$20.85M
Q3 2025	2	CodeKarma, Lightbase	$5.09M
Q4 2025	2	Nozomio, CoreStory	$38.20M
Q1 2026	1	Adronite	$5.00M
Q2 2026	1	GitHits	$1.75M
Full period	8	8 companies	$70.89M
What we see	Persistent deal flow	Several independent teams	Mostly small-to-mid-sized rounds

Did Code Metal kick off a real funding wave for AI code refactoring?

Code Metal had a remarkable funding acceleration, but we still have only one company behind the entire Refactoring AI Tools surge.

Code Metal raised $36.5 million in Q4 2025 and followed it with a $125 million Series B the next quarter. Few companies in our sample moved from a Series A of that size to another nine-figure round so quickly.

The company also had a concrete commercial story behind the financing. At the Series A, Code Metal said its technology was already deployed across defense, automotive and semiconductor customers and was contracted to deliver eight figures of revenue during the year. By the Series B, the company was describing accelerating adoption in mission-critical industries, with customers including RTX, Toshiba, L3Harris and the U.S. Air Force.

That sales environment is very different from a general-purpose coding copilot. Translating and verifying software for defense hardware, vehicles or semiconductor systems can involve large contracts and a much higher cost of failure.

The category still needs another company or two to follow a similar path before this looks like a wider funding wave. Right now, Code Metal itself is scaling fast; the category around it is much less proven.

Why are corporate investors suddenly showing up in AI coding?

Corporate investors started showing up when AI coding companies became directly relevant to their own technology stacks, customers and industries.

We found no strategic-backed deals in Q2 or Q3 2025. In Q4, NVIDIA and Google participated in Cursor's Series D, while RTX Ventures and Bosch Ventures joined Code Metal's Series A. RTX then participated again when Code Metal raised its Series B.

The connection is fairly straightforward. Google and NVIDIA sit directly inside the AI model and compute ecosystem used by coding tools. RTX and Bosch operate in industries where Code Metal sells code translation and verification technology.

Strategic-backed companies accounted for only two of five Q4 deals and one of two Q1 deals, yet those financings represented nearly all the capital raised in both quarters.

Corporate capital appears most interested once an AI coding company has reached serious commercial or strategic relevance. None of the Seed financings in our five-quarter sample included a strategic investor.

Quarter	Strategic-backed deals	Share of deals	Share of funding
Q2 2025	0	0%	0%
Q3 2025	0	0%	0%
Q4 2025	2	40%	98.1%
Q1 2026	1	50%	96.2%
Q2 2026	0	0%	0%
Main companies	Cursor, Code Metal	-	-
Corporate names	NVIDIA, Google, RTX, Bosch	-	-

Are the same VCs taking over AI coding funding?

AI coding funding is still spread across a fairly wide investor set, with only a few firms appearing repeatedly.

Accel has the clearest recurring presence, showing up as a disclosed lead or co-lead on three financings during the wider dataset. LocalGlobe appears twice. Most other disclosed lead investors appear once.

Q4 was particularly broad. Five deals involved ten disclosed lead or co-lead names, although several rounds had syndicates and we should not interpret each name as a completely separate lead decision.

The much shorter investor lists in Q1 and Q2 2026 mostly reflect the drop to two deals and then one deal. We do not see a clear progression where the same two or three venture firms gradually capture the entire market.

Is North America taking over AI code assistant funding?

North America dominates AI coding scale capital, while the smaller startup pipeline remains much more international.

Every Q4 2025 and Q1 2026 deal in our sample came from North America. Those were also the quarters containing Cursor and Code Metal, so the region captured virtually all of the large-round capital.

The pattern becomes less extreme when we look at young companies. Q3 had one North American company, one European company and one Asia-Pacific company. Q2 2026 then consisted entirely of GitHits, a European startup.

With only one to five deals per quarter, geographic shares can change dramatically because of a single financing. The cleanest read is that the biggest AI coding companies raising growth capital are heavily concentrated in North America, while new-company activity still appears elsewhere.

Why did AI coding funding dry up so badly in H1 2026?

AI coding funding became extremely thin in H1 2026 because the wider group of companies raising meaningful rounds in Q4 disappeared almost immediately.

Q1 produced only two financings. Code Metal carried almost the whole quarter, while Adronite raised $5 million. The broad mix seen three months earlier, two $30 million-plus rounds, several smaller financings and five funded companies, was gone.

Q2 went further. GitHits' $1.75 million Seed round was the only qualifying transaction we found.

Even removing the large Code Metal financing from Q1 leaves $5 million, so the weakness cannot be dismissed as another comparison distorted by a megaround. Funding underneath the outliers genuinely dried up.

One quarter this small still needs some caution. With only a single deal in Q2, percentages such as 100% Seed, 100% Europe or 100% first financing mostly describe GitHits, not a stable market pattern.

Is most of the big AI coding money going to companies that have already proved themselves?

Large AI coding rounds are increasingly reserved for companies that can already show revenue, enterprise adoption or strategic importance.

The contrast with first financings is striking. New startups continued to get funded throughout the period, yet their combined capital was tiny beside the rounds going to established companies.

Cursor is the clearest example, with hundreds of millions in recurring revenue before its largest financings. Code Metal moved quickly because it could point to mission-critical deployments, contracted revenue and recognizable industrial and government customers.

That creates a two-speed market. Investors are still placing relatively cheap bets on new AI coding ideas, while the largest pools of money go to companies where commercial proof already exists.

For founders, the difficult jump seems to be the space between those two groups: moving from a small Seed or Series A into a broad, repeatable market for $50 million-plus rounds.

Why does AI coding funding swing so wildly every quarter?

AI coding funding swings so wildly because there are very few large companies in the market, and their financing schedules can completely reshape a quarter.

A single round supplied more than 96% of funding in Q2 2025, Q4 2025 and Q1 2026. In a dataset with only one to five deals per quarter, moving one of those transactions by a few weeks would radically alter the chart.

Deal counts are much calmer. They moved from four to three to five to two to one. Headline capital, meanwhile, moved through hundreds of millions and billions of dollars.

The middle of the market gives us a much better read. Funding below $50 million improved through late 2025 before dropping sharply in 2026. Category breadth and first-financing activity also continued even when headline funding collapsed.

So when someone says AI coding funding “grew 7,000%” or “fell 95%” from one quarter to another, the first thing to check is which company raised.

Did the Q2 2026 AI coding funding drought continue after the quarter ended?

The Q2 2026 drought was severe, but the latest evidence says large AI coding capital did come back after our measured period.

Our five-quarter analysis ends on June 30, so later transactions are excluded from every calculation above. That boundary is important because Q2 itself contained only the $1.75 million GitHits round.

The picture changed afterward. The Wall Street Journal reported that NVIDIA agreed to invest $1 billion in Poolside as part of a much larger technology and licensing transaction. Poolside builds AI models and tools around software development, putting the transaction directly back at the scale end of the AI coding market.

That post-period deal does not rescue Q2 statistically, and one huge financing still says very little about the breadth of the market. It does change the current interpretation: Q2 was a bad quarter, but the evidence available today does not support treating it as the start of a lasting freeze in large AI coding financings.

Which AI code assistant funding trend actually looks real today?

The clearest AI code assistant funding trend today is a growing gap between frequent small startup bets and rare, extremely large financings for companies that have already reached scale.

Code Understanding Tools give us the best evidence that new company formation is healthy enough to persist. Eight companies in that category were funded during the five-quarter period, and at least one appeared every quarter.

At the opposite end, Cursor and Code Metal show how quickly capital can concentrate once a company demonstrates unusual revenue growth, enterprise adoption or strategic value. The later Poolside transaction suggests investors are still willing to commit very large sums to AI coding companies after the weak Q2 period.

Several other patterns are much less settled. Corporate investors appeared strongly for two quarters and then disappeared from Q2. North American dominance weakened as soon as the large U.S. rounds vanished. Refactoring AI looked hot because Code Metal was hot.

The shape of the market is the part worth taking most seriously: plenty of experimentation at the bottom, a thin middle, and enormous checks available to the handful of companies that manage to break through.

What is the funding trend actually telling us?

AI code assistant funding is still active, but calling it a broad funding boom would miss what is happening underneath. Q4 2025 showed that several companies could raise meaningful rounds at once, yet that breadth disappeared in H1 2026.

The most persistent activity is happening among smaller companies. Code understanding, repository context and tools that make AI-generated software more reliable continue to produce new startups and financings.

At the top end, capital is remarkably concentrated. A small number of companies with clear commercial traction can raise hundreds of millions or even billions, making the entire market appear much larger from one quarter to the next.

We also see a high bar for moving from early funding into scale capital. Seed and Series A rounds keep happening, while comparatively few companies make the jump into very large follow-on financings.

Corporate investors are worth watching. Their appearance around Cursor and Code Metal suggests AI coding has become strategically important to model providers, chip companies and industrial groups, although two active quarters are still too little history to call it permanent.

Finally, Q2 2026 should be remembered as a genuine funding drought within our measured period, not as proof that investors had lost interest in AI coding. The post-period Poolside financing shows that nine- and ten-figure capital is still available today when investors see a company they believe can operate at scale.