Tech Digest – February 17, 2026
Defense, Autonomy & AI Governance
Pentagon Pits SpaceX Against Defense Primes on Autonomous Drone Swarms — While Threatening to Blacklist Anthropic
SpaceX and its subsidiary xAI are competing in a secretive $100 million Pentagon prize challenge to develop voice-controlled autonomous drone swarming technology. xAI also quietly released Grok 4.20 Beta with multi-agent reasoning this week. Meanwhile, the Department of Defense is reportedly threatening to cut all ties with Anthropic and designate it a “supply chain risk” for attempting to restrict military applications of its models on classified networks.
Note: Two signals, one conclusion. The defense establishment wants AI without usage restrictions. Companies that draw ethical lines around military deployment risk losing access to the largest single buyer on earth. Every AI vendor serving institutional clients is now implicitly choosing a side.
The Memory Crisis Reshapes Hardware Markets
Nvidia Skips a GPU Generation for the First Time in 30 Years — Sony, PC Makers, and Consumers Pay the Price
Nvidia will not release any new gaming GPUs in 2026 — the first time in nearly three decades the company has skipped a full year. The reason: a global DRAM shortage driven by AI data center demand, with AI chips commanding 65% profit margins versus 40% for gaming. Sony is considering delaying the next PlayStation to 2028 or 2029 as memory shortages squeeze supply. Across Europe’s five largest markets, refurbished PC sales are climbing 7% as new devices become unaffordable. Micron’s CEO has confirmed memory markets will “remain tight past 2026.”
Note: This isn’t a product delay. It’s a structural reallocation: AI gets the silicon, everyone else gets the leftovers. Any institution planning hardware procurement in the next 18 months should budget for longer lead times and higher prices — or pivot to refurbished.
Sources: Tom’s Hardware, Yahoo Finance / Bloomberg, The Register
Infrastructure & Capital Flows
India Becomes the Next AI Infrastructure Frontier — $100 Billion in Data Centers, Hundreds of Millions in VC
Adani Group announced plans to invest $100 billion in renewable-powered AI data centers across India by 2035. Separately, major venture capital firms including Khosla Ventures, Accel, and Lightspeed are each committing $300 to $500 million to India’s AI ecosystem in a coordinated capital push.
Note: The compute buildout is going global. India joins the US, the Gulf states, and Southeast Asia in the race to host AI infrastructure. Europe’s relative absence from these announcements is becoming conspicuous.
Sources: Reuters, Economic Times
Small English Towns Fight Back Against Data Center Sprawl
Communities in rural England are protesting plans to convert agricultural land and forest into server halls. The resistance highlights a growing tension between national AI infrastructure ambitions and local land use, environmental concerns, and community opposition.
Note: The pattern will repeat across the EU. Data centers need power, water, and land — all of which are governed locally. Municipalities that don’t have a position on this yet will find one forced on them.
Sources: Wired
Scientific Automation
“AI Will Do to Physics in 2026 What It Did to Coding in 2025”
Alex Lupsasca, a coauthor of OpenAI’s recent paper on gluon physics, says AI is about to transform scientific research the way it transformed software development last year. A fellow physicist noted that the problem AI solved — computing gluon scattering amplitudes — had long been considered “an elaborate way of arriving at zero.” The bottleneck wasn’t intelligence. It was human attention. The calculation was possible but too tedious for anyone to prioritize.
Note: The “boring but important” backlog exists in every institution. When AI clears the attention bottleneck in physics, the same logic applies to compliance reviews, procurement evaluations, and regulatory analysis — tasks that are well-defined but never get done because nobody has the bandwidth.
Sources: Alex Lupsasca (X), Martin Bauer (X)
Smallest Self-Replicating Molecule Found — 45 Nucleotides in Ice
Researchers have discovered the first small polymerase capable of self-replication, composed of only 45 nucleotides and operating in mildly alkaline eutectic ice. The finding, published in Science, sheds new light on how life may have originated — and represents another milestone in AI-accelerated laboratory science.
Sources: Science
Robotics & Manufacturing
Chinese Humanoids Sprint at 35 km/h and Perform Autonomous Kung Fu — Robotics Enters a New Phase
MirrorMe Technology unveiled Bolt, a full-size humanoid robot (175 cm, 75 kg) that hit 35 km/h in real-world testing — faster than most humans will ever sprint and just seconds off Olympic pace over 100 meters. Separately, Unitree showcased dozens of G1 humanoids performing the world’s first fully autonomous robot cluster Kung Fu routine for Lunar New Year, demonstrating coordinated high-speed movement across multiple machines simultaneously. Chinese researchers also published DISH, an ultra-rapid holographic 3D printing method in Nature, fabricating millimeter-scale objects in 0.6 seconds at 19-micron resolution.
Note: Three developments from different Chinese labs in a single week: speed, coordination, and precision manufacturing. The gap between demonstration and deployment is shrinking faster than most procurement cycles can adapt.
Sources: Fox News / CyberGuy, Unitree Robotics (X), Nature
Agent Economics & Digital Trust
First Platform Charges AI Agents More Than Humans — The Pricing Split Begins
Polylogue, a collaborative writing platform, has introduced what it calls AI-discriminatory pricing: the service is free for human users, but costs $10 per month to add an AI agent to a workspace. AI agents join as real team members — reading documents, leaving comments, editing alongside humans — and get billed accordingly.
Note: If agents consume resources like users, they’ll get priced like users. Any institution deploying AI agents into shared platforms should expect this to become standard — and start budgeting for it.
Sources: Polylogue
