$800 Million in Saudi Oil Money Just Landed in Open-Source AI Infrastructure

Here's the deal: on July 1, 2026, Together AI announced it had closed an $800 million Series C. And the lead investor wasn't just any VC firm — it was Aramco Ventures, the venture capital arm of Saudi Arabia's state-owned oil giant, Saudi Aramco. That single check pushed Together AI's valuation to $8.3 billion. Rewind 16 months to its Series B, and the company was worth $3.3 billion — so this is roughly a 2.5x jump in just over a year. There's something almost poetic about a state-linked oil company, built on fossil fuel money, leading a massive round into a company whose whole business is making open-source AI models run in production. It's a pretty loud signal about where the AI industry is heading. The closed frontier-model ecosystem has had the spotlight for a while now, and the open camp is finally mounting a real counterattack.

This isn't just another "big funding round" headline, though — there's substance behind it. Together AI says it now generates more than $1 billion in annual bookings, with reports putting the figure closer to $1.15 billion. On top of that, usage of open-source models on its platform has tripled over the past year. That's not vaporware valuation inflation — that's actual companies paying actual money to run open models at scale. The announcement also named thousands of customers already running on the platform, including recognizable names like Cursor, Cognition, and Decagon.

The Players

Let's start with what Together AI actually does, because the term gets thrown around loosely. It's what people in the industry now call a "neocloud" — a new breed of AI infrastructure provider. In plain terms, it's the layer that lets enterprises train and run open-source models like DeepSeek, Nemotron, MiniMax, and Kimi, at a fraction of what it would cost to run equivalent workloads on closed models — while matching or sometimes beating their performance. Instead of every enterprise having to stand up its own GPU clusters and figure out optimization from scratch, Together AI provides an optimized software stack on top of that hardware, turning open models into something you can actually trust at enterprise scale.

The people who built this company aren't randoms, either. The cofounders are Vipul Ved Prakash, Percy Liang, Ce Zhang, and Chris Ré. Percy Liang and Chris Ré are well-known names in the Stanford AI research world, and Ce Zhang has a long research background in distributed systems and machine learning infrastructure. This is a team that came out of academia with a fairly consistent mission from day one: build the infrastructure layer that keeps the open-source AI ecosystem viable at scale.

Aramco Ventures, the lead investor here, is a very different kind of character. Saudi Aramco is the largest oil company on the planet, and Aramco Ventures is its corporate venture capital arm — the vehicle that takes oil-fueled capital and aggressively deploys it into future-facing industries, AI and energy transition being the current obsessions. The fact that a state-linked energy giant's VC arm is leading the round for an open-source AI infrastructure startup says a lot about how seriously Gulf capital is now taking AI infrastructure as an asset class, not just a side bet.

Beyond Aramco, the round pulled in a genuinely stacked list of participants. Private equity heavyweight Vista Equity Partners is in, along with early-stage AI backers General Catalyst and Emergence Capital. NVIDIA — the undisputed GPU market leader — also joined, as did March Capital, Taiwanese electronics manufacturer Pegatron, and S Ventures, the venture fund tied to cybersecurity firm SentinelOne. The presence of hardware and manufacturing players like NVIDIA and Pegatron is a tell: this round isn't just a financial bet, it looks like a supply-chain and ecosystem-level play too.

What Actually Happened

Break down the numbers and the trajectory here is pretty striking. About 16 months before this announcement, Together AI closed a $305 million Series B at a $3.3 billion valuation. Now, with $800 million more in the bank, the valuation sits at $8.3 billion — roughly a 2.5x jump in a little over a year. That kind of velocity in a valuation is unusual in venture territory. Investors generally don't reprice a company that fast unless the revenue growth is undeniable and hard to argue with.

And the revenue side genuinely backs it up. Together AI says annual bookings have now crossed $1 billion, with reporting putting the number closer to $1.15 billion. "Bookings" here matters as a word choice — it implies contracted revenue, meaning customers have actually signed on the dotted line, not just revenue projections dressed up to look impressive. Layer on top of that the tripling of open-source model usage on the platform over the past year, and you've got both revenue and usage climbing at triple-digit percentage growth simultaneously. That's the kind of combination that gives institutional investors real conviction to write a big check.

The customer roster is also worth a closer look. Together AI says it now serves thousands of customers, including AI coding tool Cursor, autonomous software engineering company Cognition, and AI customer service agent builder Decagon. What these companies have in common is that they're all running AI models in live, real-time, production environments — not research sandboxes. That tells you Together AI's infrastructure is handling actual production traffic at scale, and that's very likely the core evidence that got Aramco and the rest of the investor syndicate to open their checkbooks.

Here's a quick table breaking down the key numbers from this round.

Item Detail
Announcement date July 1, 2026
Round Series C
Amount raised $800 million
Post-money valuation $8.3 billion
Lead investor Aramco Ventures
Prior round (Series B) $305 million, $3.3 billion valuation
Valuation growth ~2.5x in 16 months
Annual bookings Over $1 billion (~$1.15B reported)
Open-source model usage growth 3x over the past year

Looking at that table, it's clear this round is more than "a deep-pocketed investor made a bet." It's underpinned by concrete, verifiable metrics on revenue and usage. And the presence of hardware supply-chain players like NVIDIA and Pegatron on the cap table suggests this is a strategic move spanning the entire GPU and infrastructure supply chain, not just a purely financial position.

What Each Side Gains

For Together AI, this round is basically a massive war chest. AI infrastructure is a fundamentally capital-intensive game — you're constantly spending money to build and maintain GPU clusters, continuously optimize open-source models, and rapidly integrate whatever new model drops next, whether that's DeepSeek, Nemotron, MiniMax, or Kimi. The $800 million gives the company the firepower to compete aggressively for GPU capacity instead of getting squeezed out, and to keep expanding infrastructure on the offensive rather than playing defense. And with the valuation now at $8.3 billion, the company also gets a meaningful edge in recruiting top talent and negotiating future funding rounds — in startup land, valuation functions as both credibility and leverage.

For Aramco, the calculus is more layered. On the surface, this looks like a straightforward financial bet on a fast-growing startup. But zoom out, and it reads more like a piece of a bigger strategy: reallocating oil wealth ahead of a post-oil future. Saudi Arabia has been trying to shift petrodollars into future-facing industries for years now, and AI infrastructure is currently one of the hottest destinations for that capital. Aramco stepping in as lead investor for an open-source AI infrastructure company signals that Gulf capital is moving past just funding "AI application layer" startups and starting to place direct bets on the infrastructure layer itself. Rather than lining up to invest in closed frontier-model labs, this looks like a deliberate strategic choice to grab a stake in the core infrastructure powering the open camp.

The other investors each have their own angle. For NVIDIA, Together AI is both a core customer and a partner that consumes massive volumes of its GPUs — the bigger Together AI gets, the more NVIDIA chips get sold, making this a textbook ecosystem investment. For a manufacturer like Pegatron, this likely represents a move to cement its footing in the AI infrastructure hardware supply chain. And purely financial investors like Vista Equity Partners, General Catalyst, and Emergence Capital are making a classic growth-stage bet — riding a company whose revenue and usage are both compounding at triple-digit rates, with an eye toward a big payoff at a future IPO or acquisition.

Lastly, this is good news for the enterprise customers already depending on Together AI — companies like Cursor, Cognition, and Decagon. Their core infrastructure partner just got financially much more stable, which naturally raises confidence in service reliability and the long-term product roadmap. If the infrastructure that lets you run open-source models cheaply keeps growing and getting backed by this much capital, there's less and less reason for these companies to feel locked into expensive closed models by default.

Precedents: Wins and Failures

Open-source software infrastructure companies backed by huge capital have absolutely succeeded before — Red Hat is the classic example. Built on top of the free, open-source Linux operating system, Red Hat made its money by wrapping support, services, and reliability guarantees around it so enterprises could trust it in production. That business ultimately got acquired by IBM for $34 billion. It proved a formula: the open-source core is free, but the layer that makes it trustworthy and usable for enterprises is where the real money lives. What Together AI is doing right now looks like the AI-era version of exactly that formula. Models like DeepSeek and Kimi are released for free, but the value concentrates in the infrastructure layer that lets enterprises run them reliably, cheaply, and safely.

Cloud infrastructure startups backed by big capital have also seen rapid growth before. Data infrastructure companies like Snowflake and Databricks raised large rounds and saw their valuations multiply several times over in just a few years — Databricks eventually crossed a $10 billion-plus valuation. What these companies share is that they secured a position as essential infrastructure that developers and data engineers use every single day. Together AI appears to be chasing a similar position: becoming the mandatory gateway that any company wanting to put AI models into production has to go through.

That said, it's worth being honest about the failure cases too. Plenty of AI infrastructure and cloud startups have struggled after flashy funding rounds, undone by overinvestment and misjudged demand forecasts. GPU clusters in particular depreciate fast and carry enormous maintenance costs — if a company overcommits on those assets and actual customer demand doesn't materialize as expected, cash flow can deteriorate fast. There are cautionary tales from 2023-2024, when some AI cloud startups signed overly aggressive GPU leasing deals only to run into demand slowdowns. Whether Together AI's $800 million turns into durable growth or gets burned inefficiently will come down to how well the company reinvests it into GPU capacity and software optimization.

Sovereign capital making massive startup bets has a genuinely mixed track record. The most infamous example is SoftBank's Vision Fund, backed partly by Saudi and Abu Dhabi sovereign wealth, pouring enormous sums into WeWork — a bet that turned into a well-documented disaster. On the flip side, some of that same capital made early bets on AI and semiconductor companies that are performing quite well today. So whether Aramco's bet on Together AI ends up looking like a WeWork repeat or a successful infrastructure play is genuinely too early to say. What's clear, though, is that unlike WeWork's overhyped SoftBank-style bet, this investment came with actual revenue and usage metrics behind it — that's a meaningful difference.

Rivals' Counterplay

News that Together AI just secured this much dry powder is not something competitors get to shrug off comfortably. The first group that should be feeling the heat is closed frontier-model providers. Companies like OpenAI and Anthropic have largely justified their pricing with a straightforward pitch: "our models are so much better that you have no real alternative." But if platforms like Together AI keep closing the performance gap between open and closed models while undercutting them dramatically on cost, that pitch starts to lose its footing. The framing in this announcement — "comparable or better performance at a fraction of the cost of closed systems" — reads like a direct shot across the bow of the closed-model camp.

Competitors in the same neocloud and AI infrastructure category also have reason to sit up. This space includes GPU-cloud specialists like CoreWeave, as well as inference-speed-and-cost players like Fireworks AI and Groq. Together AI landing an $8.3 billion valuation with $1 billion-plus in annual bookings puts real pressure on these rivals to raise bigger and move faster. As sovereign wealth funds and strategic investors start taking open-source AI infrastructure seriously as a category, competition for follow-on funding is only going to intensify from here.

Big Tech's response is also worth watching closely. Google, Microsoft, and Amazon already offer their own services for serving open-source models on their respective clouds. For these hyperscalers, a growing Together AI represents a credible alternative competing directly with their own cloud platforms. That said, given their existing capital advantages and massive installed customer bases, they're unlikely to feel immediate existential pressure — the more probable response is doubling down on integrating open-source model serving more aggressively into their own platforms rather than treating this as a five-alarm fire.

On the flip side, this deal could actually be read as a green light for other open-source AI infrastructure startups. When a mega-investor like Aramco puts real conviction behind an entire category, it tends to validate that category for the broader market — "open AI infrastructure is a legitimate place to put capital" becomes a more credible narrative. Big rounds tend to create a trickle-down effect where capital starts flowing more freely to other startups in the same sector, which is a pattern venture capital has seen play out many times before. So Together AI's round is simultaneously a competitive threat and, in a way, free marketing that grows the entire open-source AI infrastructure category.

So What Changes

For AI startup founders, the implication here is pretty clear. The question of "should we build on open-source models or closed models" now has a much more credible answer on the open side. A company like Together AI landing this much capital and continuing to grow signals that the infrastructure for serving open models reliably is going to keep improving and getting cheaper over time. For an early-stage startup thinking through its cost structure, treating open-model infrastructure as a serious default option — rather than an afterthought — is starting to look like the rational move.

For developers, more options is just good news, full stop. As infrastructure for reliably running open-source models like DeepSeek, Nemotron, MiniMax, and Kimi in production gets more robust, the path to building a genuinely good AI product without being locked into a single closed API widens considerably. That's especially welcome for developers who've been worried about vendor lock-in. That said, "open" doesn't automatically mean "better" — closed frontier models still hold a real edge on plenty of hard, specialized tasks, and that's worth keeping in mind.

For investors, this deal is likely to become something of a benchmark going forward. When evaluating AI infrastructure startups, "does this company show triple-digit growth in both revenue and usage simultaneously" is probably going to become a more central criterion for setting valuations. And the fact that sovereign capital like Aramco directly bet on open-source AI infrastructure could be read as an early signal that similar sovereign wealth funds and state-linked capital pools will get more aggressive about entering the AI infrastructure sector. Whether that actually plays out remains to be seen, but interest in this sector has clearly stepped up a notch.

For everyday users and enterprise decision-makers, the end result of this trend is likely to show up as lower costs for using AI, full stop. As open-model infrastructure gets cheaper and keeps scaling reliably, the barrier to entry for companies wanting to bolt AI features onto their products drops, and that competition tends to eventually flow back to end users in the form of better pricing or improved service quality. None of this changes overnight, of course — how much of this actually materializes still depends heavily on how efficiently Together AI deploys the capital it just raised into real infrastructure improvements.

🥄 Three Things You're Probably Wondering

— Is Aramco actually serious about AI, or is this just parking money somewhere? On the surface it looks like a straightforward financial bet, but there's a strong case that it's tied into Saudi Arabia's broader post-oil strategy. Whether Aramco keeps making follow-on bets in AI infrastructure or this ends up being a one-off symbolic move is too early to say.

— Is this $8.3 billion valuation just hype? It's hard to call it pure hype when it's backed by over $1 billion in annual bookings and 3x usage growth. That said, how well those bookings convert into actual cash revenue, and whether this growth rate holds up, are things we'll need more time to see play out.

— Are closed frontier-model companies actually in trouble now? Not an immediate existential crisis — closed models still hold a real edge on plenty of complex, specialized tasks. But the shift of cost-conscious enterprise customers toward the open camp is a trend that's becoming genuinely hard to ignore at this point.

References

Numbers and criteria are as of announcement and may change. Investment calls are yours to make!