Reflection AI: From $545 Million to $27.5 Billion in Under a Year

Reflection AI was valued at $545 million in mid-2025. By October 2025, after a $2 billion round led by Nvidia with $800 million committed, it was at $8 billion. By March 2026, after closing a $2.5 billion round, it reached a $27.5 billion post-money valuation. The company was founded in 2024 and has roughly 60 employees. The trajectory tells you less about what Reflection AI has built and more about what the market is willing to pay for strategic positioning in open-weight AI infrastructure before it has been demonstrated at scale.

The Company

Reflection AI was founded by Misha Laskin and Ioannis Antonoglou, both former Google DeepMind researchers, to build open-source AI systems positioned as the Western answer to DeepSeek. Its primary product is Asimov, a code-research agent designed to help engineering teams understand large, complex codebases, rather than generate new code. Code understanding at scale is a harder and more defensible problem than code generation, where competition from GitHub Copilot, Cursor, and others is intense, and that product choice is central to the company's differentiation argument.

The open-weight strategy is the other pillar. In collaboration with Nvidia, Reflection releases open-source AI frameworks accessible to enterprises, governments, and academic institutions, with sovereign AI partnerships with US allied nations as a stated priority.

Why the DeepSeek Context Matters

To understand why Reflection AI is valued at $27.5 billion, you need to understand what DeepSeek R1 did to the market in January 2025. When DeepSeek released R1 on January 27, 2025, claiming performance comparable to OpenAI's o1 at a fraction of the cost, the Nasdaq Composite fell 3.1% as Nvidia dropped 17% and lost $589 billion in market capitalisation—the largest single-day loss in US stock market history. The model was open-source, cheap to deploy, and came from a Chinese lab, three things the US AI establishment had assumed would not arrive simultaneously for years.

Before R1, China's AI industry was largely centred on closed models. Open models existed but were mostly confined to research communities. R1 changed that. Within months, Chinese open-weight models were dominating benchmark leaderboards and Western AI communities were scrambling for commercially deployable alternatives that did not carry data sovereignty concerns or geopolitical exposure. That is the market Reflection AI is explicitly building for, and it is why JPMorgan's Security and Resiliency Initiative, a $10 billion programme focused on economic stability and national security, is writing checks into a 60-person startup.

Who Is Backing It

The investor roster across Reflection's rounds tells the strategic story clearly:

• Nvidia — $800 million committed, securing chip demand and embedding Nvidia's products at the foundational layer of Reflection's infrastructure
• JPMorgan — Security and Resiliency Initiative, national security-adjacent capital treating open-weight AI as systemic infrastructure
• Sequoia Capital and Lightspeed — traditional institutional venture from the companies' earliest rounds
• Reid Hoffman, Alexandr Wang, Databricks Ventures — operator and strategic capital with direct AI deployment experience

The $2.5 billion March 2026 round included JPMorgan through its Security and Resiliency Initiative, a $10 billion programme the bank built to invest in companies critical to economic stability and national security. JPMorgan's participation through this vehicle rather than through traditional venture channels signals that major financial institutions are treating open-weight AI infrastructure as a matter of systemic importance, with the same logic that puts defence procurement and grid infrastructure in a different investment category from speculative technology bets.

The Valuation in Context

At a $27.5 billion post-money valuation against a company of approximately 60 employees with a product still in early commercial stages, investors are paying for strategic positioning and option value rather than current revenue. That framing is consistent with how the top AI companies have been priced across the board in 2026: OpenAI at $852 billion, Anthropic at $965 billion, Moonshot at $35 billion and rising, all priced on trajectory and strategic importance rather than current profitability. Reflection at $27.5 billion is a smaller version of the same thesis, at an earlier stage with more uncertainty in the outcome.

The comparison that anchors the investment case is DeepSeek, which Reflection explicitly positions against. DeepSeek operates as a subsidiary of High-Flyer Capital and is not independently valued, but its R1 model demonstrated what commercial and strategic value a widely adopted open-weight model can generate. China's National AI Industry Investment Fund, which anchored Moonshot's recent $3.5 billion round, is the same sovereign vehicle that backs DeepSeek. Reflection is the American counterpart play, and US sovereign-adjacent capital is treating it accordingly.

The valuation velocity ($545 million to $27.5 billion in under 12 months) is less unusual in context than the raw number suggests. The AI private market has repriced faster than any technology category in history, driven by sovereign capital, strategic competition, and the belief that the companies building foundational infrastructure will capture durable positions regardless of near-term revenue. Reflection's trajectory is extreme even within that context, but it is a difference of degree rather than kind.

What the Speed of Repricing Signals

When JPMorgan's Security and Resiliency Initiative is writing checks into 60-person AI startups, the capital is following strategic positioning rather than revenue, and the belief that open-weight AI infrastructure will become as consequential as semiconductor supply chains. At 100x or 200x revenue multiples, the companies that fail to deliver on that positioning will reprice sharply. The ones that build the adoption base and the developer ecosystem that makes switching costly will compound.

That is the bet Reflection's investors are making, and the DeepSeek precedent is their evidence that it can play out. Our piece on the Wellington midyear outlook covers the entry price question in private markets in detail.

NonPublic Pty Ltd (ABN 49 607 216 928) holds Australian Financial Services Licence #482668. Investments are available to wholesale and sophisticated investors as defined under the Corporations Act 2001. This content is general in nature and does not constitute financial product advice. It does not take into account your objectives, financial situation, or needs. Investing in private markets involves significant risk, including the potential loss of your entire investment. Past performance is not a reliable indicator of future results. You should obtain independent financial advice before making any investment decision.