The AI Thesis
A weekly AI newsletter covering five key developments reshaping how we work, live, and compete. Five Minutes. No jargon. Verified facts.
STORIES FROM THE WEEK ENDING 30 OCTOBER 2026
✦ Editor's Note
The biggest bets in AI increasingly rest on someone else's computers, legal shield or balance sheet.
The Pentagon ran short of secure computing; Britain is spending £165m to avoid that kind of dependence. Delaware may let owners shelter behind an AI agent. Meta wants shareholders to fund what its cash flow no longer covers, and Broadcom is guaranteeing the loans behind its biggest customer's chips. Monday's question: which AI dependency are we carrying that is not yet on our risk register?
1
The Pentagon's First Full-Scale AI Targeting Campaign Used Up All Its Secure Computing
The Pentagon's chief digital and AI officer, Cameron Stanley, said the US assault on Iran, Operation Epic Fury, was the first full rollout of the Maven AI targeting and analysis system. Its use almost immediately consumed all available US classified computing capacity. A military official told a defence-industry conference that such models can produce errors, or "hallucinations", and must be retrained every 90 days to recognise targets such as tanks reliably. Washington is helping allies adopt similar tools, though some worry about relying on systems the US could switch off.
Why It Matters
If the world's best-funded military ran short of secure computing, private firms planning sensitive AI workloads should not assume capacity will be there when needed. Reliance on systems another country controls is now a board-level question for allies and multinationals alike.
2
Britain Puts £165m Behind Defences Against AI-Enabled Biological Threats
The Chancellor, John Healey, announced a £165 million National AI Biosecurity Initiative in a written statement to Parliament on 26 October. It includes £100 million for an AI Biosecurity Lab and £65 million to strengthen biosurveillance and develop medical countermeasures. The government argued that AI's economic benefits will only be realised if critical infrastructure, public services and businesses are resilient to emerging threats. It added that Britain must not be left vulnerable or dependent on other countries. The statement was made in both Houses of Parliament.
Why It Matters
AI-enabled biological risk is moving from research topic to national-security budget line. Life-sciences, healthcare and critical-infrastructure firms should watch how the new lab is staffed and procured, and whether its work leads to new screening or reporting expectations.
Source: UK Parliament, Written Ministerial Statement HLWS424, 26 October 2026
3
Delaware's Plan for Companies Run by AI Agents Draws a Congressional Investigation
Delaware, which dominates US company incorporation, is drafting rules for "Artificial Intelligence Companies": entities run day to day by autonomous AI agents that could sign contracts and sue or be sued, inside a 30-month regulatory sandbox. Owners would keep limited liability if the company is adequately capitalised and its activities logged. Banking would be off-limits. On 26 October, James Comer, the Republican chair of the House Oversight Committee, demanded documents from Governor Matt Meyer by 9 November, warning the plan could shield AI firms from responsibility. A bill may follow in January.
Why It Matters
Giving AI agents corporate powers would shift liability from people to capital buffers and activity logs. General counsel, insurers and boards should start asking who answers when an agent, theirs or a counterparty's, signs a bad contract.
Source: Financial Times, 26 October 2026 · House Oversight Committee, 26 October 2026
4
Meta Asks Shareholders for Up to $95bn to Keep Building AI
On 26 October, Meta launched about $35 billion of share offerings: roughly $22.5 billion of Class A stock sold through forward agreements and $12.5 billion of mandatory convertible preferred shares. It also plans an at-the-market programme of about $60 billion once its new Class C shares begin trading. The company said the money supports continued investment in AI infrastructure, and it expects 2026 capital spending of $135–145 billion. In the third quarter, capital spending of $34.9 billion exceeded the $32.8 billion of cash generated by its operations.
Why It Matters
When one of the world's most profitable companies turns to equity markets, the AI build-out is no longer self-funding even at the top. Expect more equity-linked issuance across the sector, dilution debates and sharper investor scrutiny of returns on AI spending.
5
A Chipmaker Guarantees the Loans That Let Its Biggest Customer Lease Its Chips
Bank of America, Citi, Morgan Stanley and Blackstone will provide $63 billion to a new special purpose vehicle that will buy Broadcom's AI chips and networking equipment and lease them to Anthropic. Broadcom is providing "residual value support" to the senior lenders, agreeing to absorb falls in the hardware's value up to the size of the first-lien debt. Moody's called Broadcom one of the strongest large technology companies but warned that risk would rise if several such deals followed in quick succession. The financing was announced on 26 October, the day Anthropic filed for its stock-market listing.
Why It Matters
A supplier is now underwriting its customer's hardware risk. Lenders, lessors and any business buying AI capacity on long contracts should ask who absorbs the loss if today's chips lose value faster than expected, and how concentrated that exposure has become.
Source: Joint release via Broadcom Investor Relations, 26 October 2026 · Reuters via Khaleej Times, 27 October 2026
🌿 AI Making A Difference: Good News
Routine Health Records Could Help Spot People at Higher Risk of Common Cancers
Researchers built around 500 machine-learning models to predict the risk of 20 common cancers, using blood tests, medical history and lifestyle data from 463,620 UK Biobank participants followed for about 18 years. Results were most promising for lung, oesophageal and liver cancers. Presented at the ESMO Congress on 26 October, the approach could help target screening, though prospective studies must confirm it improves outcomes.
Source: European Society for Medical Oncology via EurekAlert, 26 October 2026