Nvidia in talks to acquire, invest in or acqui-hire Reflection AI
The chipmaker is exploring multiple paths to deepen its relationship with the open-weight model developer, which already counts Nvidia as its largest backer.
Nvidia Corp. is negotiating with open-weight model developer Reflection AI Inc. over a potential acquisition, additional investment or acqui-hire arrangement, according to reporting by the Financial Times. Sources with knowledge of the discussions indicated to the publication that a deal could materialize within weeks, though negotiations remain fragile and could still fall apart.
The Financial Times outlined several possible structures for how the transaction might proceed. Nvidia could acquire Reflection outright, boost its existing stake—potentially bundled with expanded chip and computing capacity—or pursue an acqui-hire model. Under an acqui-hire structure, Nvidia would bring Reflection's team on board and gain access to its technology without assuming ownership of the company itself. The FT noted that this approach would likely sidestep a lengthy regulatory approval process.
Nvidia has employed both strategies in recent months. In December, inference chip startup Groq Inc. licensed its technology to Nvidia in a transaction valued at $20 billion. Founder and Chief Executive Jonathan Ross and other top executives joined Nvidia, while Groq continued operating its cloud business independently. Hugging Face Inc. underwent a straightforward acquisition when Nvidia agreed on Sept. 3 to pay $12.9 billion for the model hosting platform.
Reflection already maintains a substantial financial relationship with Nvidia. The chipmaker led Reflection's $2 billion funding round in October 2025 and has invested $800 million in the company overall. Chief Executive Misha Laskin told CNBC in April that Reflection's most recent round closed at a $25 billion pre-money valuation. Laskin and Ioannis Antonoglou, who both previously worked as researchers at Google DeepMind, established the company in 2024.
The Financial Times story surfaced five days after Reflection introduced Beam, its inaugural open-weight model. Beam operates as a mixture-of-experts architecture containing 501 billion parameters. The company initially concentrated on coding agents and is positioning Beam as a competitor to lower-cost Chinese open models including DeepSeek and Kimi for coding and agentic applications.
Reflection trained Beam using Nvidia GB300 systems leased from Space Exploration Technologies Corp. under a June agreement with a potential value of up to $6.3 billion. Monthly payments of $150 million secure access to capacity at SpaceX's Colossus 2 data center in Memphis through 2029. Nebius Group N.V. also committed in July to provide Reflection with an additional $1 billion in AI computing capacity.
Reflection operates as a venture-backed startup with backing from Sequoia Capital Operations, Lightspeed Venture Partners LP and DST Global LP alongside Nvidia. Neither Nvidia nor Reflection has made any public statements regarding the ongoing discussions.
Read more
Speed to operational status now matters more than GPU counts in AI infrastructure buildouts
As organizations race to deploy AI factories, the bottleneck has shifted from securing graphics processors to managing data pipelines and storage. Infrastructure vendors are rethinking how they approach these projects from the ground up.
SpaceXSI's Grok Voice 3.0 Tops Industry Benchmarks in Speech-to-Speech Showdown
Elon Musk's SpaceXSI has unveiled Grok Voice 3.0, a speech-to-speech model that outperforms competitors from OpenAI and Google across multiple independent testing frameworks.
Microsoft Pursues Composable Memory to Ease AI Data Center Bottlenecks
As memory scarcity constrains AI workloads across data centers, Microsoft is exploring composable memory architectures that dynamically allocate capacity across servers rather than locking it within individual machines.