Top 5 Tech News — Monday, October 19, 2026

Top 5 Tech News — Monday, October 19, 2026

Welcome to today's tech roundup. The US Department of Homeland Security ordered federal civilian agencies to strip DeepSeek from their systems, AI chip startup Tenstorrent raised a $1.5 billion round at a $6.2 billion valuation, Anthropic pledged 5% of its shares to a new public benefit fund, Microsoft walked away from the UK's flagship sovereign AI datacentre over power shortages, and Meta detailed its million-GPU Prometheus supercluster. Here are the top five tech stories from the last 24 hours.

1. DHS Orders DeepSeek Out of Federal Civilian Agencies, Including Tools Built on Its Models

The Department of Homeland Security has ordered all US federal civilian agencies to stop buying and immediately remove technology from Chinese AI company DeepSeek — including third-party products "developed using DeepSeek implementations." The Exclusion and Removal Order, signed July 24 but made public via the Federal Register on October 19, warns DeepSeek's models present "multiple threat vectors," including the risk of sensitive data transferring to China and malware embedded in the models.

Why it matters: This is the first government-wide ban on a frontier AI model, and it reaches into any vendor product built on DeepSeek's open weights — a hazy scope DHS hasn't fully defined. The order signals that US-China AI decoupling is moving from chip export controls to software itself.

2. Tenstorrent Raises $1.5 Billion at $6.2 Billion Valuation for Open-Weight AI Chips

Nvidia challenger Tenstorrent has raised $1.5 billion in a Series E round valuing the company at $6.2 billion post-money, more than doubling its December 2024 valuation. The Santa Clara startup, led by chip architect Jim Keller, will use the funding to scale production of its Galaxy servers for serving open-weight AI models and video generation. The round was led by Trace Capital and Innovation X, with ARK Invest, SanDisk, Morgan Stanley, and T. Rowe Price participating.

Why it matters: Five months after reportedly fielding acquisition interest from Intel and Qualcomm, Tenstorrent chose independence — a sign investors believe there's real money in chips built specifically for open models rather than Nvidia's one-size-fits-all stack.

3. Anthropic Commits 5% of Its Shares to a New Public Benefit Fund

Anthropic announced on October 18 that it will commit 5% of its shares to a new Anthropic Public Benefit Fund, subject to stockholder approval, with the board able to add 2% more for every 50% increase in the company's value. The fund will be a wholly owned LLC, and none of its equity may be used for lobbying or political giving. Combined with existing co-founder and employee pledges, roughly 20% of Anthropic's shares will now be dedicated to public-benefit purposes.

Why it matters: As the AI industry faces what CEO Dario Amodei calls "a crisis of trust," this is the most concrete attempt yet by a frontier lab to tie its own financial upside to public benefit — though investors will watch closely how the "direct distributions" to workers displaced by AI actually materialize.

4. Microsoft Pulls Out of UK's Flagship Sovereign AI Datacentre Over Power Supply Delays

Microsoft has told developer Nscale it is ending its agreement to anchor the planned AI datacentre at Loughton, Essex, after grid operator UK Power Networks said it could not supply enough electricity to the site until the early to mid-2030s. The Guardian, which broke the story on October 18, reports the project — hailed by ministers as the UK's largest sovereign AI datacentre — has lost its only publicly announced customer. Microsoft remains involved in a separate Nscale project in Norway.

Why it matters: Power, not chips, is becoming the binding constraint on the AI buildout: even a flagship government-backed project got shelved over grid timelines measured in decades. The exit also complicates Nscale's planned Wall Street listing, reportedly targeting a $35 billion valuation.

5. Meta Brings Prometheus Supercluster's First Phase Online, Scaling Toward 1 Million GPUs

Meta published a detailed engineering account on October 19 of Prometheus, its AI supercluster in New Albany, Ohio, confirming the completed first phase — hundreds of thousands of GPUs operating as a single machine — trained the Muse family of foundational models, including Muse Wonder. The gigawatt-scale campus uses high-capacity Ethernet and closed-loop liquid cooling, and Meta reports over 85% goodput during pre-training, with plans to scale toward roughly one million GPUs.

Why it matters: Gigawatt-scale AI training is no longer a press-release ambition; Meta is naming the reliability figures (85% goodput, ~47k GPU hours between interruptions) that determine whether these megaclusters actually pay off. The numbers set a new benchmark the rest of the industry will be measured against.

Sources