Two former Google DeepMind researchers have closed a $2 billion deal as their company pushes on with its plans to build frontier open intelligence.
The startup is gaining traction and its investors include Nvidia, Disruptive, DST, 1789, B Capital, Lightspeed, Eric Yuan, Eric Schmidt, Citi, Sequoia, and CRV.
Founded seven months ago by Misha Laskin and Ioannis Antonoglou, Reflection AI jumped 15 times in valuation to $8 billion in record time. Laskin led reward modeling for DeepMind’s Gemini project, and Antonoglou co-created AlphaGo, the system that defeated the world Go champion in 2016.
“We need to build open models so capable that they become the obvious choice for users and developers worldwide, ensuring the foundation of intelligence remains open and accessible rather than controlled by a few,” Reflection AI said in its announcement.
The DeepSeek wake-up call
The urgency behind Reflection’s record raise is obvious if we look at the bigger picture. Chinese companies like DeepSeek have revolutionized AI development with cost-effective models, followed by similar moves from China-based competitors like Qwen and Kimi. The tempo is fast, and the price tags are low.
While Western labs typically spend billions training frontier models, DeepSeek claims its R1 model required less than $6 million in training costs, a sliver of what companies like OpenAI invest.
CEO Laskin did not mince words about the stakes.
“DeepSeek and Qwen and all these models are our wake-up call because if we don’t do anything about it, then effectively, the global standard of intelligence will be built by someone else,” he said. “It won’t be built by America.”
The threat runs deeper than benchmarks. It touches geopolitical influence in the digital age. Chinese models promise strong capabilities at breakthrough prices, and while enterprises and governments often steer clear over potential legal complications, that hesitation creates an opening for Western alternatives that can match their efficiency.
Even Washington is taking notice. White House AI and Crypto Czar David Sacks endorsed the announcement, posting that it is great to see more American open source AI models; given global demand for cost, customizability, and control that open source offers.
The future of AI development
Reflection’s approach signals a shift in American AI strategy, away from closed, proprietary models and toward open-source competition with Chinese alternatives. The startup has built a large-scale LLM and reinforcement learning platform capable of training massive mixture-of-experts models at frontier scale. First up, autonomous coding. Then broader reasoning.
The company’s 60-person team has already secured computing infrastructure and plans to release a frontier language model next year trained on tens of trillions of tokens. Revenue will come from large enterprises building products on top of their models and governments developing sovereign AI systems, tapping demand for domestically controlled AI infrastructure.
Reflection’s definition of “open” mirrors strategies from Meta with Llama or Mistral, centering on access rather than development transparency. They will release model weights for public use while keeping datasets and training pipelines proprietary, a balance between openness and competitive advantage.
The broader market context makes this round even more striking. AI agent startups secured $2.8 billion in the first half of 2025 alone, with development tools commanding premium valuations of 30 to 50 times revenue, a clear vote of confidence in AI infrastructure.
Not everyone has good news to share about AI. Former Google CEO Eric Schmidt warned that there’s evidence AI models can be manipulated to “learn how to kill someone.”
The post Ex-DeepMind Staff Raise $2B to Challenge China’s AI Edge appeared first on eWEEK.
No Responses