By Lina Irawan 
Last week, the AI world watched closely as Demis Hassabis, the longtime CEO and co-founder of Google DeepMind, announced his decision to step back from the day-to-day operational leadership role. Hassabis, a visionary often credited with steering DeepMind to its groundbreaking achievements, including AlphaGo and AlphaFold, will transition into a new chair role. This move, while seemingly a promotion, represents a significant shift in the company’s leadership structure and potentially its foundational culture. For a co-founder who has been at the helm since DeepMind’s inception, stepping away from the CEO position often indicates a strategic re-evaluation or a response to mounting pressures.
In a move that further underscores the restructuring, Hassabis ceded immediate operational control to DeepMind’s chief technology officer, Koray Kavukcuoglu. What’s particularly noteworthy is the altered reporting line: Kavukcuoglu will now report directly to Sundar Pichai, CEO of Alphabet (Google’s parent company), rather than holding a standalone DeepMind chief title. This structural change is not merely administrative; it suggests a tighter integration of DeepMind into the broader Google organizational hierarchy, potentially eroding the distinct autonomy that DeepMind has historically enjoyed. Just minutes after Hassabis’s announcement, another seismic departure rocked the AI community: chief scientist Jeff Dean, a legendary figure within Google and a titan in the field of AI and systems engineering, also announced his exit. Dean’s departure, following closely on the heels of Hassabis’s transition, amplifies concerns about the loss of institutional knowledge and scientific leadership at a critical juncture for Google’s AI ambitions. These high-profile exits, far from isolated incidents, appear to be just the tip of the iceberg in what has been a tough year for DeepMind.
The challenges extend beyond leadership. DeepMind’s product development has hit significant roadblocks. Gemini 3.5 Pro, a highly anticipated large language model unveiled with considerable fanfare at Google I/O in May, has now blown through three separate release dates. Such delays are not merely inconvenient; in the fast-paced world of AI, they can severely impact market perception, developer adoption, and competitive positioning. While Google showcases its innovation at major events, the inability to deliver promised models on schedule raises questions about internal execution, resource allocation, and the complexity of developing state-of-the-art AI.
Adding to the product woes, independent benchmarking company Artificial Analysis recently provided a stark assessment to Fortune, revealing that Google’s best shipped model, Gemini 3.6 Flash, currently trails a host of prominent rivals. The analysis indicates that Gemini 3.6 Flash lags behind models from Anthropic, OpenAI, xAI, Meta, and even at least one Chinese lab in terms of "raw intelligence." This finding represents a significant reversal from a brief but memorable stretch last year when Google actually topped the AI leaderboard, showcasing its formidable capabilities. The rapid shifts in the AI competitive landscape highlight the intense pace of innovation and the difficulty of maintaining a consistent lead. "Raw intelligence" in this context refers to a model’s fundamental reasoning capabilities, problem-solving prowess, and general understanding, which are critical for developing truly advanced AI applications. Falling behind on this core metric is a serious concern for a company that has positioned itself as an AI-first leader.
Multiple engineers within DeepMind have voiced concerns that the recent reshuffle compounds a slow-motion pull of power from London, where DeepMind was famously founded, to Mountain View, where Google’s sprawling headquarters are located. This geographical and organizational shift is deeply symbolic and strategically significant. For years, a "firewall" had been maintained between DeepMind and its parent company, Google. This firewall was designed to protect DeepMind’s unique research-first culture, allowing it to pursue ambitious, long-term scientific endeavors without being immediately constrained by Google’s product cycles or commercial imperatives. Engineers worry that losing Hassabis, the architect and guardian of this autonomy, means this longstanding firewall is now breaking down. Such a breakdown could lead to DeepMind’s research agenda being more directly dictated by Google’s immediate product needs, potentially stifling the kind of fundamental, curiosity-driven research that led to its most significant breakthroughs. While a Google spokesperson disputes this framing, asserting that London remains central and that DeepMind retains its research autonomy under the new structure, the sentiment among engineers on the ground tells a different story. The fear is that DeepMind might evolve from a pioneering research lab into a more integrated product development arm of Google, potentially diminishing its distinct identity and innovative edge.
The company is also struggling to retain some of its key talent, a critical asset in the knowledge-intensive field of AI. In a single, particularly challenging week in June, Google experienced a severe brain drain. Noam Shazeer, a co-lead on the Gemini project, departed for OpenAI, one of Google’s primary competitors. Even more impactful was the loss of John Jumper, a co-inventor of the revolutionary AlphaFold, who moved to Anthropic. AlphaFold, which accurately predicts protein structures, was a monumental scientific achievement that transformed biological research and earned Jumper a Nobel Prize. The exodus didn’t stop there, as two more AlphaFold veterans, Jonas Adler and Alexander Pritzel, followed Jumper out the door. Engineers attribute this significant talent drain to a confluence of factors: aggressive poaching from cash-flush rivals like OpenAI and Anthropic, who are offering lucrative packages and compelling research environments; growing frustration within DeepMind over falling behind on critical coding benchmarks and product releases; and the allure of pre-IPO equity at these younger, rapidly scaling rivals, promising substantial future payouts. The loss of such foundational figures, especially those behind breakthroughs like AlphaFold, is a severe blow to DeepMind’s research capabilities and its ability to innovate at the cutting edge.
None of these challenges, however, mean that Google is out of the AI race. Far from it. Google still possesses immense structural advantages that none of its AI-native rivals can currently match. It commands unparalleled resources in custom silicon chips (such as its Tensor Processing Units, or TPUs), vast cloud infrastructure through Google Cloud, and an immense global distribution network spanning Android, Search, Chrome, and YouTube. These fundamental advantages in hardware, infrastructure, and user reach could potentially offset a weaker showing at the model layer. The ability to deploy models at scale, gather vast amounts of data, and leverage an integrated ecosystem provides a robust foundation that smaller, AI-native companies simply cannot replicate overnight.
However, just four years after Google’s first major "AI code red"—a period in the late 2010s when the company recognized the urgent need to reorient itself around AI—DeepMind is once again trying to prove it can move fast enough to keep pace with an accelerating industry. This time, the challenge is compounded by the absence of the two figures most closely associated with its scientific identity and strategic vision: Demis Hassabis and Jeff Dean. Their departure marks the end of an era for DeepMind and signals a new chapter, one that will likely see greater integration with the broader Google ecosystem, potentially prioritizing product development and commercialization over pure, unconstrained research. The coming months will be crucial in determining whether DeepMind can navigate these tumultuous waters and reassert its dominance in the global AI landscape, or if its foundational shift will lead to a different, less pioneering role within the tech giant.
Beyond the internal shifts at tech giants, the broader investment landscape continues to buzz with activity, reflecting sustained confidence in AI and other burgeoning sectors.
VENTURE CAPITAL
The venture capital scene remains robust, with significant funding rounds for companies across various high-growth sectors. Thrive Holdings, a New York City-based AI holding company that strategically acquires professional-services businesses, successfully raised a substantial $2 billion. This massive round saw participation from major investors including D1 Capital Partners, Altimeter Capital, and SoftBank Group, underscoring strong investor belief in the AI-driven consolidation model. In a testament to the surging interest in personal AI, River AI, a Palo Alto, Calif.-based developer of innovative personal AI tools, secured an impressive $1.1 billion in funding. The round was co-led by General Catalyst and AMP PBC, with key strategic investments from technology giants NVIDIA and AMD Ventures, highlighting the industry’s commitment to advancing AI at the individual user level.
Defense technology continues to attract significant capital, as evidenced by Heaviside Industries, a Los Angeles, Calif.-based company focused on developing autonomous munitions, which raised $60 million in Series B funding. Felicis led this round, joined by a consortium of investors including Hedosophia, Menlo Ventures, Cantos, Flume Ventures, Qstar Capital, Friends & Family Capital, Anorak Ventures, and MVP Ventures. In the fintech space, Yuno, a New York City-based payments and financial-services software company, closed a $45 million Series B round. Global PayTech Ventures spearheaded the investment, with follow-on participation from prominent firms like a16z, Tiger Global, QuantumLight, Monashees, and Kaszek.
The advertising technology sector is also seeing innovation, with Gravity, a San Francisco-based ad network specifically designed for AI products, raising $38 million across its seed and Series A funding rounds. Caffeinated Capital led the seed investment, while Lightspeed Venture Partners took the lead in the Series A funding. In the realm of AI-powered creative tools, Preview, a San Francisco-based AI-video production software company, secured $12 million in funding. Sequoia led this round, joined by The General Partnership. Furthermore, Genera, a San Francisco-based AI deployment software company, raised $10 million in seed funding, with First Round Capital leading the investment and support from BoxGroup, Wndr, Carpenter Capital, and Success Venture Partners. Finally, in cybersecurity, Cytix, a Manchester, U.K.-based software company, raised $7 million in Series A funding, led by Northern Gritstone, reflecting ongoing demand for robust digital protection.
PRIVATE EQUITY
Private equity firms continue to execute strategic acquisitions and recapitalizations across diverse industries. Astorg announced its acquisition of the global microbiology business from Thermo Fisher Scientific, a Waltham, Mass.-based life sciences products and services company, with financial terms remaining undisclosed. This move indicates a focus on specialized segments within the life sciences sector. Atlas Merchant Capital led a minority investment in Edge Focus, a New York City-based consumer-credit technology company, also with undisclosed financial terms, signaling confidence in the evolving credit technology market.
Blue Point Capital Partners recapitalized Dumpsters.com, a Westlake, Ohio-based dumpster-rental and waste-services company, reinforcing the trend of private equity investment in essential services. Goldman Sachs made a notable move by agreeing to acquire NEOS Investments, a Westport, Conn.-based ETF platform, underscoring the growing interest in the exchange-traded fund market. Marlin Equity Partners acquired Cemplicity, an Auckland, Australia-based patient-experience software company, highlighting the demand for specialized software solutions in healthcare. Lastly, New Heritage Capital acquired a majority stake in PowerRail, an Exeter, Penn.-based aftermarket locomotive-parts company, showcasing investment in critical industrial infrastructure.
FUNDS + FUNDS OF FUNDS
In the fundraising landscape, Team8, a New York City and Tel Aviv, Israel-based venture capital firm, successfully raised $365 million for its third fund. This fund is specifically targeted at Seed- and Series A-stage founders building AI-native companies across critical sectors such as cybersecurity, software infrastructure, fintech, and digital health, reflecting a strategic focus on foundational AI innovation.
PEOPLE
The industry also saw key leadership appointments. Northleaf Capital Partners, a Toronto, Ontario-based private equity firm, announced the hiring of Michelle Joliat as its new Chief Operating Officer. Joliat brings extensive experience to the role, having previously served with the prestigious Ontario Teachers’ Pension Plan.
See you tomorrow,
Beatrice Nolan
X: @beafreyanolan
Email: [email protected]
Submit a deal for the Term Sheet newsletter here.
Joey Abrams curated the deals section of today’s newsletter. Subscribe here.
