Safe Superintelligence
A $32B “straight-shot” AI research lab with no product and no revenue — how NVIDIA and Alphabet’s compute bets are underwriting a conviction-priced asset
Safe Superintelligence Inc. (SSI) was founded in June 2024 — one month after Ilya Sutskever’s departure as OpenAI’s Chief Scientist. Headquartered in Palo Alto, California, with a second research site in Tel Aviv, the company operates as a deliberately small organization (an estimated ~50 employees). SSI positions itself as a “straight-shot” lab pursuing the single objective of safe superintelligence, with no interim commercial products or revenue targets. Since inception, the company has disclosed almost nothing about its research for nearly two years, maintaining an unusually strict information-security posture even by frontier-lab standards.
Born in 1986 in Nizhny Novgorod, in the former Soviet Union, Sutskever emigrated to Jerusalem, Israel at age five and later relocated to Canada in 2002. He earned a bachelor’s degree in mathematics at the University of Toronto, followed by a master’s and PhD in computer science under Geoffrey Hinton — widely regarded as a father of deep learning. After completing his PhD in 2013, he spent a brief postdoctoral stint with Andrew Ng at Stanford before joining Google Brain, after Hinton’s startup DNNResearch was acquired by Google. He is a co-creator of AlexNet (2012), the convolutional neural network credited with catalyzing the modern deep-learning era, and at Google Brain co-developed the sequence-to-sequence learning framework with Oriol Vinyals and Quoc Le — foundational work that underpins today’s transformer and GPT-class models. He is also a co-author of the seminal AlphaGo paper. In 2015, Sutskever co-founded OpenAI alongside Sam Altman, Elon Musk, and others, serving as Chief Scientist and steering research that led to the GPT model series and, later, OpenAI’s o1-class reasoning models. He was among the OpenAI board members involved in the November 2023 ouster of Sam Altman, and departed OpenAI in May 2024 — founding SSI one month later. He assumed the CEO role in July 2025 following the departure of co-founder Daniel Gross.
A former OpenAI researcher and SSI co-founder. Following Daniel Gross’s June 2025 departure, Levy was promoted to President and now co-leads the organization alongside Sutskever, with the technical team continuing to report directly to Sutskever.
Formerly head of AI at Apple, and co-manager (with Nat Friedman) of the investment partnership NFDG. As SSI’s founding CEO, Gross led early organization-building, but departed for Meta Superintelligence Labs on June 29, 2025, shortly after Meta’s attempted acquisition of SSI was rebuffed by Sutskever. Meta separately acquired up to a 49% stake in NFDG.
SSI does not operate a “business” in the conventional sense. There is no disclosed product, demo, revenue, or public roadmap. The company’s ability to raise capital rests almost entirely on the scientific credibility of Sutskever and the core research team, together with Sutskever’s own public assertion that SSI’s research has reached a stage “worthy of scaling.” Given this structure, an assessment of SSI’s operating status must center on compute access and strategic-partnership architecture rather than conventional product or revenue metrics.
The organization is structured around a single objective — safe superintelligence — with almost no disclosure of research direction over roughly two years. Coinciding with the July 2026 NVIDIA partnership announcement, the company offered its first public indication that it has reached research results it considers “worthy of scaling.”
Alphabet participated as a strategic investor while Google Cloud concurrently began supplying SSI’s research TPU infrastructure — a dual investor/vendor relationship that mirrors arrangements seen with other frontier labs such as Anthropic, and introduces a degree of hyperscaler infrastructure dependency.
Grants SSI access to NVIDIA’s next-generation Vera Rubin platform, with the stated goal of expanding compute capacity by “an order of magnitude.” The two companies will also collaborate on technical advancement of NVIDIA’s current and future compute platforms.
Structural Peculiarity of the Business Model: SSI’s valuation has been established with no conventional revenue stream — no hardware or software sales, subscriptions, or API billing. All capital raised to date has reportedly come in as cash (not cloud credits), deployed almost entirely toward compute acquisition and recruitment of top-tier research talent. This places SSI outside standard revenue-based valuation frameworks and closer to a long-duration call option on frontier AI research.
2025 Governance Reshuffle: In June 2025, Meta attempted to acquire SSI outright; Sutskever declined, and Meta pivoted to hiring founding CEO Daniel Gross instead. Sutskever subsequently assumed the CEO role, with Daniel Levy promoted to President. The company maintains that its technical organization continued reporting directly to Sutskever throughout, preserving research continuity, but the episode nonetheless created a brief leadership gap on the commercial and operational side that merits continued monitoring as a governance risk factor.
Since its founding, SSI has raised approximately $7B cumulatively over roughly two years, with its valuation climbing more than sixfold — from $5B at the first priced round to $32B most recently. Notably, this valuation trajectory was established entirely without product or revenue validation, relying instead on founder credibility and strategic hyperscaler conviction — an unusually conviction-dependent capital-raising path even within the AI sector. No down-round has been recorded to date.
SSI was incorporated one month after Sutskever’s departure from OpenAI. A small initial capital raise (recorded as a corporate minority round) followed shortly after founding, though the specific amount was not disclosed.
NFDG, the investment partnership run by Nat Friedman and Daniel Gross, led the round, with participation from Andreessen Horowitz, Sequoia Capital, DST Global, and SV Angel. The round was funded entirely in cash (no cloud credits), with proceeds directed toward compute acquisition and recruitment of top-tier researchers. The round has been characterized as one of the highest-valued Series A raises on record.
Greenoaks Capital led with a reported commitment of approximately $500M, joined by Andreessen Horowitz, Lightspeed Venture Partners, and DST Global as follow-on investors. Notably, Alphabet and NVIDIA joined as new strategic investors, establishing a dual-track hyperscaler relationship combining equity investment with infrastructure supply. A separate TPU infrastructure agreement with Google Cloud was signed around the same time. With valuation rising more than sixfold in roughly seven months absent any product or revenue, the round stands out as a case where investor conviction moved well ahead of conventional market consensus.
Shortly after Sutskever rejected Meta’s attempt to acquire SSI outright, founding CEO Daniel Gross departed for Meta Superintelligence Labs. Meta separately acquired up to a 49% stake in NFDG, the investment vehicle run by Gross and Nat Friedman. Sutskever assumed the CEO role and Daniel Levy was promoted to President as the company reorganized its leadership.
Deal Overview: NVIDIA and SSI jointly announced a long-term strategic partnership. The official press release describes only a “substantial investment,” without disclosing a specific figure. Bloomberg, Reuters, and other outlets, citing unnamed sources, have reported the investment at approximately $5B. This figure remains unconfirmed by either company and, according to reporting, may be contingent on milestone achievement — meaning the ultimately realized amount could differ from current press estimates.
Deal Structure: Alongside the capital commitment, SSI gains access to NVIDIA’s next-generation Vera Rubin systems, with a stated goal of expanding compute capacity by an order of magnitude. The two companies will also collaborate on technical advancement of NVIDIA’s current and future compute platforms. NVIDIA CEO Jensen Huang cited Sutskever’s research lineage dating back to AlexNet in explaining the rationale for the partnership.
Implications: The deal brings SSI’s cumulative funding to approximately $7B, with the $32B valuation reported to be holding steady (it remains unconfirmed whether this constitutes a new priced round). For NVIDIA, the transaction can be read as a strategy of using compute supply as a vehicle for equity exposure — simultaneously expanding its GPU demand base and securing early access to frontier AI research capability.
SSI’s investment appeal rests not on conventional financial metrics but on the founder’s scientific credibility, a research structure insulated from commercial pressure, and compute access secured simultaneously from two competing hyperscalers. These strengths function less as a traditional economic moat and more as conviction-based assets, and should be weighed alongside the risk factors detailed in the following section.
Sutskever’s research lineage — running directly through AlexNet, sequence-to-sequence learning, AlphaGo, the GPT model series, and OpenAI’s o1-class reasoning research — represents a credibility asset held by only a handful of individuals globally. Continued public endorsement from peers such as Yoshua Bengio further reinforces this standing. In effect, this functions as human-capital collateral in the capital markets, underpinning continued fundraising in the absence of revenue.
By foregoing interim product launches and revenue targets, SSI can direct its entire research roadmap toward the single objective of safe superintelligence — a degree of research autonomy that compares favorably against OpenAI, Anthropic, and Google DeepMind, all of which operate under product and revenue pressure. This comes at the cost of a near-total absence of commercial validation metrics.
Google Cloud (TPU) and NVIDIA (Vera Rubin GPU) — direct competitors in compute supply — both serve simultaneously as equity investors and infrastructure partners. This structurally mitigates single-vendor dependency risk while giving SSI unusually broad access to top-tier compute supply chains.
Across two priced rounds, SSI has raised its valuation more than sixfold with no down-round, and the $32B valuation is reported to have held through the 2026 NVIDIA transaction. This suggests the investor base continues to treat SSI as a premium asset and supports the case for continued negotiating leverage in future rounds.
The investment thesis for SSI relies heavily on conviction-based positioning that sits outside conventional valuation frameworks. The risk factors below represent key variables in assessing the potential for downside valuation adjustment.
The $32B valuation is unsupported by any revenue, user base, or product validation metric. Absent a commercializable research breakthrough, the valuation remains highly exposed to shifts in investor sentiment or a broader tightening of AI-sector capital markets.
The June 2025 departure of founding CEO and co-founder Daniel Gross for Meta illustrates the extreme dependency risk inherent in an organization of roughly 50 people concentrated around a small number of core researchers. Against a backdrop of intensifying AI talent competition — including reported compensation packages exceeding $100M — further departures cannot be ruled out.
Key deal terms, including the reported $5B NVIDIA investment, rest largely on anonymously sourced press reporting rather than official company disclosure. While SSI attributes its deliberate opacity to research-security concerns, this creates an information asymmetry that investors must factor into valuation judgments.
The structure in which compute suppliers such as Alphabet and NVIDIA are simultaneously major shareholders raises the possibility of long-term technical or commercial dependency on specific vendor platforms, or tension between SSI’s “no commercial product” principle and strategic investors’ commercial expectations. Meta’s attempted acquisition already demonstrates that Big Tech’s commercial interest in SSI has moved beyond the hypothetical.
Competitive Landscape Context: SSI competes for talent and compute against OpenAI, Anthropic, Google DeepMind, xAI, and Meta Superintelligence Labs (which Daniel Gross has since joined) — rivals with substantially greater capital and, in most cases, existing products and distribution. Given that many of these competitors already generate revenue, whether SSI’s no-commercialization strategy ultimately converts into durable research advantage remains the central open variable in the investment thesis.