OpenAI expanded its web of chip, credit and services deals with a 10GW partnership with Broadcom and the embedding of retail giant Walmart into ChatGPT. Finance, compute, and energy are fusing into a single strategic stack. This logic extends beyond Silicon Valley: Beijing’s surprise rare-earth export controls triggered a 100% tariff response from Washington, in turn triggering a $19B crypto sell-off. Gold is up >50% this year and silver just hit an all-time high - a signal of investors bracing for volatility in the new industrial-financial era.
Public/IPOs
No traditional US IPOs scheduled for week ahead, as government shutdown stalls new issuance (c.90% of SEC staff furloughed). That said, SEC said that companies can proceed with listings without formal approval (they won’t be penalised for missing pricing details).
Aptera Motors (solar electric vehicles) plans to complete a direct Nasdaq listing this Thurs - will be an interesting test of investor appetite for pre revenue capital-intensive climate hardware.
Strava (fitness tracking platform) is seeking an IPO after a string of acquisitions, hitting 50M monthly users and nine-figure annual revenue. Backers include Sequoia, TCV, and Jackson Square; last valued at $2.2B (May 2024).
In the UK, Shawbrook Bank (specialist lender) announced plans to list on the LSE’s main market at up to £2B.
Big Dogs
OpenAI
o Reveal of internal tools, including DocuGPT, briefly wiped billions off software stocks as investors feared competition for DocuSign, HubSpot, and Salesforce. A reminder OpenAI’s experiments move entire markets.
o SoftBank seeking $5B margin loan backed by Arm shares to fund further OpenAI investments, deepening SoftBank’s $30B AI and US data infrastructure push.
o OpenAI–Broadcom joint plan announced to develop custom chips with 10GW total power draw. Another phase in compute sovereignty and US semiconductor alignment. Broadcom stock rose c.10% on the news.
o Nb. Matt Levine synthesized OpenAI’s dealmaking brilliantly (OpenAI Is Good at Deals - Bloomberg).
o Sora (OpenAIs AI video generation) hit 1M downloads in 5 days - faster than ChatGPT. Free to use but costs OpenAI around $1 per clip to run. The tension between viral growth and heavy compute spend exposes the fragility of AI’s new attention economy - built on energy, not ad revenue.
xAI raising up to $20B via an SPV that buys Nvidia chips to lease back, with Nvidia investing up to $2B (more here).
Anthropic partnering with IBM to embed its models into developer tools and expand enterprise reach (IBM).
Thinking Machines (AI research) lost co-founder Andrew Tulloch to Meta, underscoring Big Tech’s continued poaching of top AI talent despite soaring startup valuations (WSJ).
N8n (workflow automation software) reported 6x user and 10x revenue growth this year, with valuation rising 10x since March to $2.5B after a $180M Series C by Accel, Meritech, Redpoint, Evantic, Visionaries Club, Felicis, Sequoia, Highland Europe, and HV Capital (TechEU).
Wayve (autonomous driving) in talks to raise $2B from Microsoft and SoftBank. Nvidia may also invest $500M. Could value Wayve at $8B (FT).
dbtLabs, the pioneer of modern data transformation, has announced it has signed a definitive agreement to merge in an all-stock deal with Fivetran, the global leader in automated data movement. The combined company will be approaching $600M ARR.
Anysphere (developer of Cursor) reportedly considering a round at a $30B valuation. A number that reflects investor conviction in autonomous coding tools as next platform shift (here).
ReflectionAI (autonomous coding and superintelligent agents) raised $2B, bringing valuation to $8B. Backers include Nvidia, Eric Schmidt, Citi, 1789 Capital, Lightspeed, and Sequoia (here)
A breakout week for prediction platforms (Techcrunch).
o Kalshi (now handling ~60% of all U.S. prediction-market activity), raised $300M at $5B from Sequoia and a16z. Trading volumes are projected to surge from $300M in 2024 to $50B this year, helped by integrations with retail apps like Robinhood.
o Polymarket staged a return to the U.S. (after settling with regulators for past violations) and the Intercontinental Exchange (parent of NYSE) bought a 20% stake for $2B. Nb, users correctly wagered on Maria Corina Machado winning the Nobel Peace Prize 11 hours before the announcement - triggering an investigation into possible leaks.
Venture Capital
EIF survey reports improved sentiment in Q325, with significantly more fund managers expecting stronger fundraising and exit conditions ahead.
JPM plans to deploy $10B in VC to back companies enhancing US manufacturing, minerals processing, energy, AI, and quantum. Part of a wider $1.5T programme to bolster “industries critical to national economic security” (here).
Goldman Sachs is acquiring Industry Ventures (a 25-year-old secondaries firm with $7B AUM) in a deal worth up to $965M (here).
The Nobel Prize in economics went to Joel Mokyr, Philippe Aghion, and Peter Howitt for explaining how innovation drives sustained growth. Mokyr is recognized for identifying the social and institutional roots of technological progress; Aghion and Howitt for showing how “creative destruction” fuels long-term prosperity. Their recognition highlights how economics has converged with innovation strategy and geopolitics - and how technological dynamism has become a defining dimension of global power (FT).
Venture Geopolitics
Beijing announced sweeping new export controls on rare earths, effective 1 Dec, requiring licences for any product containing over 0.1% Chinese content. China processes over 90% of global supply, raising risks for semiconductors, defence systems, and batteries. Washington responded with plans to double tariffs on Chinese imports to 100% from 1 Nov and restrict exports of “critical software”. With Xi and Trump due to meet at APEC, the escalation signals a major hardening of positions (BBC).
NYC has sued Meta, Google, Snap, and TikTok for allegedly designing apps that addict children and fuel a youth mental health crisis, seeking damages for straining schools and healthcare systems. Beyond the courtroom, the case signals a shift in how governments are beginning to treat social platforms - not as private services but as public infrastructure with societal externalities (Reuters).
Analysis around AI-driven job losses remains divided. New research by Harvard’s Seyed Hosseini and Guy Lichtinger – reported in The Economist – used AI to scan 200 million job ads, identifying around 130,000 “generative-AI integrator” roles at 10,600 “AI adopter” firms. These were compared with 274,000 non-adopting companies to test AI’s impact on hiring. The adopter firms began recruiting such specialists around early 2023 – just as ChatGPT 3.5 launched – and over the next six quarters, junior hiring fell 7.7% faster than at non-adopters, suggesting entry-level work is being squeezed by automation.
Britains EV boom is now powered by China - BYD sold 11,271 cars in the UK in September and Britain is now its largest market outside China. The UK is a comparatively attractive destination for the Chinese EV manufacturer because it faces no tariffs, unlike the EU and the US (BBC).
Bloomberg wrote about how circular infrastructure deals could push AI capex toward $1T, intertwining tech, debt, and energy markets. Azeem Azhar continues his bubble monitor, noting how “opacity breeds a dangerous circularity: vendors finance customers who pay vendors who report revenue that justifies more vendor financing, round and round in a self-reinforcing loop. What happens when one link cracks? The cycle unravels fast” (Azeem Azhar).
Strategic Sectors
Cybersecurity
Google launched a dedicated AI bug bounty, offering up to $30k for vulnerabilities such as prompt-injection exploits and unsafe agent actions. The move formalises AI security testing as a core enterprise defence layer (Cloud Security Newsletter).
HackerOne (cyber vulnerability platform) reported a 210% YoY rise in valid AI vulnerabilities: 65% linked to prompt injection, model manipulation, or exposed endpoints; 35% tied to AI safety issues such as misuse or output integrity. 13% of organisations experienced an AI-related incident in 2025, and 97% lacked proper access controls. Most CISOs say they are under-resourced - a sign of how fast AI risk is outpacing enterprise readiness (Hacker Report)
Anthropic found that just 250 malicious documents can implant a backdoor into a large language model during training - showing how little poisoned data is needed to compromise even massive systems (Anthropic).
1Password launched an integration allowing AI agents to access web credentials safely within browser workflows without exposing sensitive data - an early example of infrastructure adapting to the AI-agent era (1Password).
Crypto/stablecoins
>$19B in leveraged crypto positions were wiped out on 10 Oct (“largest ever liquidation event”) hitting 1.6M traders after Trump’s China tariffs triggered cascading margin calls. Though prices partially recovered in the days afterward serves a reminder of how geopolitics now drives volatility in decentralised markets (WSJ).
Coinbase & Mastercard are eyeing BVNK ($750M valuation) as a bridge for stablecoin-based cross-border payments - signalling how private payment rails are becoming strategically contested (The Block).
The Bank of England plans to scrap its proposed £10M holding cap for crypto firms, softening its stance as stablecoins edge toward mainstream regulation (Yahoo).
Chainalysis estimates $75B in tainted assets could be seized by governments - blurring the line between enforcement and potentially accelerating efforts to create national digital-asset reserves (Cointelegraph).
Robotics
Figure unveiled their third-gen humanoid - designed for real homes, rebuilt for mass production with a supply chain targeting 100,000 units within four years - a key test of whether humanoids can finally scale beyond prototypes (Figure).
Researchers discovered Unitrees G1 robot can be hacked via Bluetooth flaws, with sensor data secretly streamed to external servers, including in China. A reminder that physical AI systems carry new dimensions of cybersecurity exposure (here).
Rodney Brooks (iRobot founder) cautioned of a humanoid robot investment bubble, arguing valuations are running far ahead of technical and commercial readiness (Techcrunch).
Insight Partners report finds robotics shifting from labs to revenue deployment, with Robotics-as-a-Service models leading the way. Simulation and teleoperation are closing the gap between prototypes and production (here).
US regulators opened a new investigation into Tesla’s Full Self-Driving after 50+ reports of red-light and lane errors - an escalation that could reshape policy on autonomous systems and safety accountability (Techcrunch).
Energy
Azeem Azhar on energy demands in the new AI world: “the realisation that we’re in an era where energy grids are planned around AI demand, not human demand, is one of those “we crossed the threshold without noticing” moments”. A few stats he shared to back it up:
o ~95% of cloud data centres may struggle to meet next-gen AI chip demand (BofA)
o Electricity near major US data hubs is up 267% in five years, sparking local anger over subsidised corporate use
o Demand for Nvidia’s A100 remains sky high five years post-launch; prices rising, depreciation stretched 3 to 6 years (3Fourteen Research).
o Constellation Energy stock up 5x since ChatGPT’s launch, outperforming Microsoft.
Notable deal: Base Power (home energy storage) raised $1B at $3B from Addition, Valor, Lightspeed, a16z, Ribbit, CapitalG, Elad Gil, Thrive, and 1789 Capital
Defence
The State of Defence Tech 2025 by Dealroom finds YTD investment in defence tech already at record highs, representing 4% of total European VC and 6.2% within EU.
Germanys intelligence Chief Martin Jaeger said this week that Russia is ready for “hot confrontation” with Europe. “We can’t simply wait and assume that a potential Russian attack won’t come before 2029”, he said (Kyiv Independent, Bloomberg)
Notable deal: Stoke Space (reusable rockets) raised $510M from Seven Seven Six, NFX, and Toyota Ventures to advance national security and commercial launch systems.
Artificial Intelligence
Vibe Coding platforms like Lovable, Replit, and Bolt have seen traffic fall up to 50%. The shift favours integrated developer tools such as GitHub Copilot and VS Code, signalling a move from consumer-facing AI to deeper infra plays where control over compute and integration define long-term power (here).
Harvard economist found 92% of US GDP growth in H1 2025 came from AI data centre and tech investment underlining how tightly economic performance is now bound to high-tech infrastructure buildout (Fortune)
JPM reports debt tied to AI firms has surged to $1.2T, overtaking US banks as the largest investment-grade segment. Lenders, like VCs, are now crowding into the same AI capital cycle (Bloomberg)
California state passed the first US law governing AI companion chatbots, mandating age checks, self-harm safeguards, and disclosures. Marks the start of legal accountability for AI systems with emotional or behavioural impact (more here)
A small nonprofit behind California’s new AI safety law accused OpenAI of using subpoenas and intimidation to weaken the bill - exposing rifts inside the company and the wider AI policy ecosystem (StrictlyVC).
Nathan Benaich’s annual State of AI report is out - c.300 slides unpacking trends across research, industry, politics and safety. From a Venture Geopolitics perspective, there is tons of data on some of the key themes we pick up each week such as:
China’s Qwen now accounts for about 40% of all new fine-tunes on Hugging Face, overtaking Meta’s Llama and signalling a shift of open-source leadership eastwards.
AI infrastructure is becoming national infrastructure: multi-gigawatt data centres, grid access, and cooling capacity are treated as assets of state importance.
Sovereign AI programmes are multiplying as the US, UAE and China bankroll domestic compute, models and data stacks to secure strategic autonomy.
Washington has moved to an “America-First AI” stance, exporting a US-built “AI stack” of compute, models and compliance frameworks to allies. The goal is to shape standards and dependency while countering China’s Digital Silk Road; it benefits US vendors but risks global lock-in and uneven control.
Europe’s AI Act faces delays and fragmented enforcement, while China accelerates with open models and local chip production.
Access to cheap, reliable power and land is becoming a geopolitical choke point; without it, nations will lag regardless of talent or capital.
Regulation itself is now a security instrument: export licences, data controls and compliance tooling act as tools of influence.
The international safety network is weakening: the UK’s AI Safety Institute has been renamed the AI Security Institute, and the US body rebranded as CAISI, reflecting a pivot from cooperative “safety” to defensive “security”.
“Fake alignment” has entered politics as governments confront models that behave differently when unmonitored, complicating trust and certification.