Tech Digest – February 10, 2026
AI Platforms & Capital
ChatGPT Now Has Ads — and 800 Million Weekly Users to Show Them To
OpenAI began testing advertisements in ChatGPT on February 9, initially targeting US users on the Free and Go tiers. The company is partnering with select advertisers for “native” ad placements within conversations. Separately, Sam Altman told staff that ChatGPT is “back to exceeding 10% monthly growth,” with 800 million weekly active users — as the company approaches a funding round that would value it near $100 billion.
Note: The free tier just became the product. Any institution that adopted ChatGPT Free as an informal productivity tool now has advertising injected into staff workflows. The business case for paid tiers — or self-hosted alternatives — just got a lot more concrete.
AI Now Claims 39% of European Venture Capital
Pitchbook data shows AI captured 39% of all European venture deal value in Q3 2025 — €17.1 billion — up from 34.5% at mid-year. For the first time, AI leads European VC funding outright. The largest rounds went to Mistral AI (€1.3 billion) and Nscale (€1.27 billion), both backed by major semiconductor and infrastructure players including ASML, NVIDIA, and Dell. Early-stage AI continues to attract capital, while follow-on rounds in fintech and life sciences are declining.
Note: When nearly four in ten venture euros go to a single technology thesis, everything else competes for the remainder. Organizations relying on non-AI digital solutions — from fintech integrations to health platforms — should watch whether their vendors’ funding pipelines are drying up.
Sources: Financial Times, Vestbee / Pitchbook
Infrastructure, Chips & Sovereignty
Alphabet Sells a 100-Year Bond to Build AI Infrastructure. The White House Wants Ground Rules.
Alphabet is issuing a 100-year sterling-denominated bond — the first century bond from a major tech company since Motorola in 1997. The issuance is part of a $20 billion-plus multi-currency debt raise to fund data center construction. Simultaneously, the White House is circulating a draft compact among tech companies to manage the resource impact of this expansion: commitments not to raise household electricity prices, to cover local infrastructure costs, and to address water consumption from cooling systems.
Note: A company betting its credit on a 100-year payoff tells you what the money thinks about AI’s staying power. The White House compact tells you what the public costs look like when the build-out lands in your municipality. Both are relevant to anyone planning infrastructure that will share the grid.
Sources: Financial Times, Politico
US Demands 40% of Taiwan’s Chip Output Relocate. Taiwan Says: Impossible.
US Commerce Secretary Howard Lutnick stated the administration’s goal is to relocate 40% of Taiwan’s semiconductor supply chain to US soil. Taiwan’s Vice Premier Cheng Li-chiun responded that this is “impossible” — the ecosystem was built over decades and cannot be transplanted by policy demand. Taiwan holds roughly 90% of the global market for advanced chips. The US has committed $250 billion in incentives plus $250 billion in credit facilities, but TSMC’s most advanced production lines remain in Taiwan.
Note: Europe’s tech sovereignty strategy assumes access to advanced semiconductors. If even a fraction of Taiwan’s output shifts to US-based fabs under bilateral pressure, EU procurement timelines and pricing for any chip-dependent project could be directly affected.
Tencent Open-Sources a 2-Bit LLM That Fits in 600MB
Tencent released HY-1.8B-2Bit, an open-source language model compressed to 2-bit precision using quantization-aware training. The result: a 1.8 billion parameter model scaled down to an effective 0.3 billion parameter footprint, requiring just 600MB of storage — smaller than many mobile apps. Despite the extreme compression, it retains chain-of-thought reasoning, runs 3–8x faster on consumer hardware (Apple M4, MediaTek Dimensity 9500), and leads equivalent-sized models by 17% on accuracy benchmarks.
Note: A capable LLM that runs entirely on a phone, with no cloud dependency, changes the data protection calculus. No API calls, no external processing, no third-party retention. For institutions exploring AI assistants in sensitive contexts, on-device deployment just became a real option — not a research project.
Sources: Tencent Hunyuan, GitHub
Workforce Signals
AI Doesn’t Reduce Work. It Intensifies It. And the Industry Building It Works 9-to-9, Six Days a Week.
A Harvard Business Review study, based on eight months of research at a 200-employee US tech firm, found that AI tools did not reduce workload — they intensified it. Employees worked faster, took on broader scope, and extended hours. Separately, BBC and Washington Post reporting documents that US AI startups are adopting “996” culture — 9am to 9pm, six days a week — driven by a perceived 2–3 year window to capture the market before Chinese competitors close the gap. China itself declared the 996 schedule illegal in 2021.
Note: The pitch for AI tools is “do more with less.” The early data says “do more with more — just faster.” Organizations deploying AI for staff productivity should measure actual workload, not just output volume. Efficiency gains that come at the cost of burnout aren’t gains.
Sources: Harvard Business Review, BBC
Autonomous Transport & Geopolitics
Driverless Cars Enter Mass Production. Connected Vehicle Software Gets a National Security Review.
Chinese autonomous vehicle company Pony AI and Toyota have begun ramping up commercial production of a self-driving car model — moving autonomous transport from pilot programs to manufacturing. Meanwhile, the US government is advancing rules to ban Chinese-developed software in connected vehicles, citing surveillance and national security risks. The auto industry is now scrambling to audit and replace Chinese code embedded in vehicle systems across the supply chain.
Note: The pattern here mirrors semiconductors: production accelerates on one side of the geopolitical line while restrictions tighten on the other. Any municipality or transport authority planning fleet modernization, smart mobility, or connected infrastructure should expect that “where the software comes from” will become a procurement question — not just “what the software does.”
Sources: Bloomberg, Wall Street Journal
And Finally
Coffee and Tea Linked to Lower Dementia Risk. Your Morning Routine May Be Doing More Than You Think.
A Harvard study published in JAMA found that moderate caffeine intake — from both coffee and caffeinated tea — is associated with a lower risk of dementia. The study tested multiple caffeine sources and found similar protective correlations across beverages. It adds to a growing body of evidence that regular, moderate caffeine consumption may have long-term cognitive benefits.
Note: After a digest full of 100-year bonds, geopolitical chip wars, and AI tools that make people work harder — it’s reassuring that the office coffee machine remains a net positive. Carry on.
Sources: JAMA
