Together AI Raised $800M From an Oil Giant. The Signal: the Money Is in the Plumbing.

News Breakdown · FiscEdge Academy

On July 1, Together AI closed an $800 million Series C at an $8.3 billion valuation — led not by a Silicon Valley fund, but by Aramco Ventures, the venture arm of the Saudi oil giant, with Nvidia, General Catalyst, Vista Equity Partners and Emergence Capital in the round.

Sixteen months ago, the same company was worth $3.3 billion. It has now roughly 2.5x'd its valuation while claiming annual bookings above $1.15 billion — renting out GPU clusters and serving open-source AI models to thousands of paying customers, including Cursor, Cognition and Decagon.

No consumer app. No chatbot. No frontier model of its own. Just infrastructure — and one of the fastest revenue ramps in enterprise history.

What Together AI actually sells

Together AI is what the industry now calls a neocloud: a specialized cloud built for one job — running AI models cheaply and fast. Its customers increasingly use it to run open-source models instead of paying for closed ones.

That detail is the story. Enterprises spent 2024–2025 paying premium prices for closed frontier models. In 2026, competent open-source models have gotten close enough in quality — at a fraction of the price — that companies are switching in volume. Together AI sits exactly on that switch, and expects to grow its compute capacity roughly 50-fold over the next five years.

An oil company leading the round completes the picture: energy players know what it looks like when the world reorganizes around a new commodity. They're betting compute is that commodity.

The founder's read: the moat moved again

Follow the sequence. First, AI made writing software dramatically cheaper — that's the whole story of vibe coding and the collapse in the cost of building a SaaS. Now the models themselves are commoditizing: open-source alternatives keep closing the gap while their price keeps falling.

When both the code and the model become commodities, where does defensibility live for a normal company?

Not in the layer everyone can rent. It lives in what you build on top: your distribution, your workflow depth, your proprietary data, your customer relationships. Together AI's own customer list proves the point — Cursor and Cognition don't win because they have secret models; they win on product and workflow, running largely on rented, open infrastructure.

For a founder, falling model prices are pure margin expansion — if you've built something customers can't easily leave. If your product is a thin wrapper around a model, the same price collapse that helps you also helps the ten competitors who launched last month.

The uncomfortable question to ask yourself

If your AI costs dropped 90% tomorrow, would your business get stronger — or would your competitors just catch up faster?

Strong answer: costs drop, your margins grow, your moat (data, workflow, distribution) stays intact. Weak answer: the model was the product, and now it's everyone's product.

If you remember one thing

The picks-and-shovels layer of AI is minting giants, and it's doing it by making intelligence cheaper for everyone above it. The winners of the next phase won't be the companies with the best model — they'll be the ones who turned cheap intelligence into products people can't leave.

---

We teach exactly this playbook — building durable products on top of commoditizing AI — in FiscEdge's AI for entrepreneurs course and the hands-on Vibe Coding Masterclass. Browse the full blog for more breakdowns. Follow @fiscedge for daily Business & AI analysis.