Together AI, one of the emerging cloud service providers that rents out NVIDIA chip servers to AI developers, is in discussions with investors to raise approximately $10 billion in funding, according to two informed sources. The pre-money valuation for this round is set at $75 billion.
This financing would more than double the company's valuation compared to one year ago and is expected to help NVIDIA diversify its customer base. Investors have expressed concerns about the chip designer's heavy reliance on a few major traditional cloud service providers.
One source indicated that Prosperity7 Ventures, the venture capital arm of Saudi oil giant Aramco, is negotiating a substantial investment. Together AI has previously raised $537 million from investors including General Catalyst, Kleiner Perkins, NVIDIA, and Lux Capital.
A spokesperson for the startup declined to comment on the current funding round.
The company initially leases NVIDIA chip servers from established cloud providers and subleases them to developers. It is increasingly purchasing its own servers, renting them out through its own data centers, following a similar model to other GPU cloud providers like CoreWeave, Nebius, and Lambda. The company also offers services for running and optimizing open-source models for developers.
NVIDIA has invested in and supplied chips to these companies to build a competitive barrier against major traditional cloud providers like Microsoft, Amazon, and Google. These giants are developing their own AI chips in an effort to reduce their cloud customers' dependence on NVIDIA hardware.
The two sources stated that Together AI's annualized revenue is approximately $10 billion, representing a more than threefold increase since mid-2025. Lambda, which plans to go public as early as this year, reported revenue exceeding $520 million for the fiscal year spanning October 2024 to September 2025.
Together AI stated that its clients include code assistant startup Cursor, AI customer service software developer Decagon, and voice AI company Cartesia.
Persistent shortages in NVIDIA chip supply mean that even traditional cloud providers need to source some chips from these emerging GPU cloud services. Together AI executives mentioned that the company also generates revenue by leasing GPUs to other cloud providers. An informed source revealed that the company has secured a multi-year, multi-billion dollar agreement with a cloud provider or a large AI enterprise, although the specific partner's identity has not been disclosed.
Competitors have secured similar deals. For instance, Microsoft leases servers from GPU cloud provider CoreWeave, while Lambda rents chips to both Microsoft and Amazon.
The new funding will provide additional capital for the four-year-old company to expand its procurement of NVIDIA chips. The company has recently been shifting from a leasing model to directly purchasing NVIDIA chips, aiming to improve its gross margin by owning the hardware rather than merely subleasing servers.
Executives stated that Together AI plans to possess 250 megawatts of proprietary data center computing capacity by the end of this year. This implies it will operate over 100,000 NVIDIA graphics processing units across multiple data centers. This capacity is roughly one-third that of its main competitor CoreWeave. Industry estimates suggest the construction cost for such capacity would be at least $8 billion.
CoreWeave, the largest among the GPU cloud providers (also referred to as new cloud vendors), is already publicly traded. Despite its soaring sales, its losses have raised investor concerns. CoreWeave reported last week that its fourth-quarter revenue doubled, but its losses widened further. Its stock price has fallen 60% from its 52-week high last June, though it remains 85% above its IPO price.
Comments