Together AI Raises $800 Million for GPU Neocloud at $8.3 Billion Valuation

Together AI, a San Francisco-based GPU cloud provider for open-source AI models, has raised $800 million in a Series C round led by Aramco Ventures, valuing the company at $8.3 billion, 2.5x its $3.3 billion valuation just 17 months earlier, one of the steeper valuation climbs among AI infrastructure companies this year. In a "neocloud" category where CoreWeave has already IPO'd after raising over $14 billion and Lambda Labs closed a $1.5 billion Series E, Together AI's round is notable less for its size than for the specific bet behind it: per-token pricing instead of the per-hour model its larger rivals use.

Last updated August 4, 2026: added real scale comparison against CoreWeave and Lambda Labs, and the per-token vs. per-hour pricing distinction that explains Together AI's actual market position, neither of which were in the original report.

What Happened: Together AI's $800M Series C

Together AI announced its Series C on July 1, 2026. The company reports annual bookings exceeding $1.15 billion, and its platform enables organizations to run proprietary and open-source large language models, including DeepSeek, MiniMax and Kimi, without purchasing GPUs directly. Founded in 2022, the company said it plans to scale computing capacity roughly 50-fold over the next 5 years with the new funding, a target that implies a substantial multi-year data-center and hardware procurement commitment well beyond what the $800 million round alone would cover, suggesting additional capital raises or infrastructure partnerships are likely still ahead.

The round drew from Vista Equity Partners, General Catalyst, Emergence Capital, NVIDIA, March Capital, Pegatron, Salesforce Ventures, Lux Capital, Geodesic, PSP Partners, and others, alongside lead investor Aramco Ventures, a broad enough investor base to span both traditional venture firms and strategic corporate backers.

Per-Token vs. Per-Hour: Together AI's Actual Differentiation

The competitive story here isn't "another neocloud raised money," it's how Together AI positions itself against the two much larger players it's compared to. CoreWeave and Lambda Labs both built their businesses on per-hour GPU rental, essentially renting out raw compute capacity. Together AI instead found its product-market fit charging per-token, based on the number of API calls, positioning itself as a developer-experience-centric layer that can run on top of infrastructure like CoreWeave's and Lambda's rather than competing purely on raw GPU access. That's a meaningfully different business, closer to an API platform than a data-center landlord, and it's the reason Together AI's smaller round doesn't mean a smaller opportunity: it's playing a different part of the stack.

Why It Matters

AI compute remains one of the industry's most persistent bottlenecks, and the scale across this category, CoreWeave's IPO and $14 billion-plus raised, Lambda's $1.5 billion round, Together AI's $800 million at a 2.5x valuation step-up, shows investors treating AI infrastructure as durable demand rather than a short-term spike. Together AI's open-source, per-token approach specifically appeals to organizations wary of vendor lock-in on proprietary APIs from OpenAI or Anthropic, without requiring them to manage raw GPU infrastructure themselves.

Per-token pricing also shifts risk in a way that matters to customers evaluating which model to use. Renting GPUs by the hour means paying for capacity whether or not it's fully utilized, which is manageable for large, sophisticated AI labs that can keep clusters busy around the clock but is a harder cost to predict for a smaller company running variable, bursty workloads. Per-token pricing ties cost directly to actual usage, which is closer to how most software services are billed and easier for a finance team to forecast against product revenue. That's part of why Together AI's positioning as a developer layer, rather than a raw-capacity provider, has found real product-market fit even against much larger-scale competitors: it's solving a cost-predictability problem, not just a compute-availability one.

What Happens Next

Together AI will need to prove its 50-fold capacity-scaling plan is achievable and that per-token pricing holds up as GPU costs and open-model competition both keep shifting. Whether its developer-layer position stays differentiated as CoreWeave and Lambda add similar API-level offerings of their own is the more interesting competitive question than the funding total itself. Both larger rivals have the balance sheets, post-IPO in CoreWeave's case, to build their own developer-facing API layers if Together AI's approach proves durable enough to be worth copying, which would turn today's complementary relationship into a more direct competitive one.

Final Takeaway

Together AI's $800 million round is smaller than CoreWeave's or Lambda's largest raises, and that's the point: it's building a different layer of the AI infrastructure stack, developer-facing and per-token, on top of the raw GPU capacity those larger players provide. The 2.5x valuation jump suggests investors believe that layer is worth a premium of its own, and Aramco Ventures leading the round is itself notable: it signals capital tied to one of the world's largest energy companies looking specifically at AI infrastructure's developer-facing layer rather than only the capital-intensive data-center buildout that has dominated most sovereign and energy-sector AI investment to date.

Key Points

The Neocloud Market

"Neocloud" describes cloud providers specialized in AI workloads rather than general-purpose infrastructure, letting them optimize hardware, software, and pricing specifically for AI training and inference. Within that category, providers split along real strategic lines rather than being interchangeable: CoreWeave and Lambda Labs compete primarily on raw GPU capacity at scale, while Together AI's open-source-model, per-token approach targets developers who want an API-level experience without directly managing that raw infrastructure. That's also Together AI's key differentiator from proprietary-model cloud AI services: customers can run open-weight models including DeepSeek, MiniMax and Kimi, which matters to organizations concerned about vendor lock-in, data privacy, or the ability to customize models for specific use cases.

FAQs