This week on Tom's Hardware Premium: October 3, 2026

Over at Tom's Hardware Premium, we've been publishing a slew of stories centered around our themed AI Chip Design Week, where you can learn more about how artificial intelligence is reshaping the chipmaking world. Each article in our series of stories is free to access for a limited time, and we're extending the free period over the weekend until Monday, October 5, so you don't miss out.

Headlining this week is our interview with OpenAI hardware VP Richard Ho, talking all things Jalapeño, the Broadcom co-developed inference ASIC that the company debuted earlier this year at Hot Chips 2026. The interview dives into how the company used AI within its design processes, beating Nvidia on efficiency, and demystifying common misconceptions like only being optimized to run OpenAI models. This lengthy read is a full, unredacted transcript of our session with Ho, so don't miss out.

• OpenAI Jalapeño design interview transcript — hardware VP Richard Ho explains how AI-assisted design may shape the future of inference ASICs

But how close is AI to designing the very silicon that it runs on? We explore the concepts, products, and everything else shaping the industry in our pair of exhaustive primers on the subject, covering the companies seeking to make the chips themselves.

We've also broken down all of the tools currently in use, with comments directly from Synopsys, and a breakdown of all claims from vendors currently using AI within their chip design workflows. Additionally, during the week, Synopsys debuted its 'Autopilot' platform for its AgentEngineer, specifically designed to aid directly in chip design workflows.

• Silicon is starting to design silicon — how AI is being used in chipmaking, from EDA tools to OpenAI's Jalapeño and beyond
• The state of agentic AI in chip design tools in 2026 — Cadence, Synopsys, and Siemens all pitch autonomous engineers
• Synopsys debuts Autopilot platform for developing chips autonomously using AI — New AgentEngineer platform is poised for 'general availability' by the end of 2026

But all of these new tools open up an uncomfortable reality. With AI's growing role in the chip design world, at what point are patented IPs in danger? This question is explored in our detailed report, where we ask experts like Domenec Forte, professor of electrical and computer engineering at the University of Florida, and Simon Moore, professor of computer engineering at the University of Cambridge, to learn more about a looming issue that could broadly impact the AI chip design market.

• AI's chipmaking frontier may face patent infringement hurdles as autonomous tools take over — 'AI can spread a copied design or infringed patent across thousands of chips before anyone notices,' says expert

With our AI chip design week in full swing, we've also been publishing our regular remit of detailed news analysis stories, exclusively for subscribers, which kicked off with a curious discovery: Intel has reportedly been granted a patent that embeds MicroLEDs in chip packaging, which could enable optical interconnects through TGVs, opening the potential for glowing RGB CPUs.

• Intel patent embeds MicroLEDs in chip packaging — technology may enable embedded optical interconnects through TGVs

Next up, blockchain-assisted cyberattacks are on the rise and have primarily originated from adversarial actors linked to Russia, North Korea, and China. The groups reportedly leverage open-weight LLMs to develop sophisticated cyberattacks against unsuspecting users.

This is a topic highlighted this week too by Anthropic, which published a report on how Zhipu AI's GLM-5.3 can be used to generate harmful content, with claims that the Chinese frontier open-weight model also has advanced hacking abilities, nearing similar levels to Claude's own Mythos AI model, showcasing different methods of getting around the model's guardrails, or disabling them entirely.

• Blockchain-assisted cyberattacks surge fivefold, driven by Iranian and North Korean state actors, Russia-linked groups — open-weight LLMs are linked to an increase in attacks
• Anthropic claims popular Chinese AI model has Mythos-class hacking abilities — frontier red teaming report details weak safeguards on open-weight AI

The AI industry continues to march on, with Nvidia releasing the Open Agent Safety Platform this week, in collaboration with hundreds of additional companies. The program aims to embed OpenShell sandboxing with BlueField-powered sentry hardware to monitor and quarantine potentially rogue agents. The move likely comes following calls for a slowdown in AI development and high-profile security concerns.

However, the costs for intelligence continue to drop, as more capable, cheap models are released on a weekly basis. While we didn't get the onslaught of model releases this week as we did last week, which saw the launch of Claude Sonnet and Opus 5.5, and OpenAI's GPT 6.1-Sol and Luna, a new report from Epoch AI claims that costs are getting just under 50% cheaper every quarter, outpacing even Moore's Law.

• Nvidia launches Open Agent Safety Platform to restrain rogue AI agents — new hardware-and-software security stack can quarantine agents in milliseconds
• The price of AI is crashing faster than the rate of Moore's Law, report suggests — intelligence costs are in freefall, outpacing comparative technologies like compute, DNA sequencing, and lithium batteries

Remember, we're keeping our AI Chip Week articles free to access over the weekend, but to read our other stories and gain access to Bench, you'll need to purchase a subscription. Your support goes a long way in helping us to publish deeper, more insightful articles that might otherwise fall outside of our regular areas of coverage.