2026-07-23 — Chips, Glasses, and Geopolitics: AI's Big Moves on July 23, 2026

Transcript

Chapter 1

Nova: Today, July 23rd, 2026: AMD drops a $5 billion bet on Anthropic, Samsung shows the world what its smart glasses actually look like, and the White House is still trying to figure out what to do about Chinese AI.

Ray: IBM's mainframe is apparently a casualty of the AI budget wars, and ServiceNow is putting $40 million on an Indian banking software firm most people haven't heard of.

Nova: Five stories, one very busy Wednesday — and the Samsung glasses reveal is the one you'll want to stick around for.

Chapter 2

Nova: The Verge AI reports AMD is committing up to $5 billion to Anthropic — and this isn't just a check. As part of the deal, Anthropic will deploy up to 2 gigawatts of AMD's Instinct MI450 AI GPUs through AMD's new Helios rack infrastructure. That's AMD planting a flag: we're not just selling chips, we're building the future of AI compute alongside the model makers.

Ray: The 'up to' is doing a lot of work in that sentence. Up to $5 billion means AMD controls the pace — they can drip this out over years and call it a $5 billion partnership on every press release. Is this a capital commitment or a very expensive co-marketing arrangement?

Nova: Even if it's staged, the MI450 GPU deployment is concrete. Two gigawatts of compute is not a press release — that's physical infrastructure going in the ground. AMD needs a flagship AI customer to challenge Nvidia's dominance, and Anthropic needs compute that isn't entirely NVIDIA-dependent. Both sides have real incentive to make this work.

Ray: Fair. For listeners tracking the chip market: this is the clearest signal yet that Anthropic is actively diversifying its hardware stack away from a single supplier. Whether AMD can actually deliver at that scale without the yield and software ecosystem Nvidia has spent years building — that's the unresolved question.

Chapter 3

Nova: TechCrunch AI has two enterprise stories worth pairing. First: ServiceNow is putting $40 million into BusinessNext, an Indian banking software specialist, at a $700 million valuation. The pitch is vertical AI dominance — get deep into banking workflows before the big incumbents even notice the category exists.

Ray: Forty million dollars at a $700 million valuation is a 5-6% stake. That's a toe in the water, not a transformation. ServiceNow gets a strategic partner; BusinessNext gets a logo and some runway. Whether that actually reshapes global banking AI is a much longer story.

Nova: The logo matters, though. ServiceNow's enterprise relationships open doors BusinessNext couldn't knock on alone. That's the real asset — distribution, not just capital.

Ray: Meanwhile, also from TechCrunch AI: IBM's stock crashed last week on weak mainframe sales, and the CEO's explanation is that AI spending ate the corporate hardware budget — temporarily, the CEO insists. That word 'temporarily' is carrying a lot of weight for a company whose mainframe revenue has been structurally declining for over a decade.

Nova: I'd push back on 'temporary' too. If enterprises are raiding hardware budgets to fund AI infrastructure, that reallocation doesn't reverse just because AI spending normalizes. The mainframe's budget competition just got a new, very aggressive rival.

Ray: The consequence for anyone in enterprise procurement: AI is already reshuffling where the dollars go, and legacy hardware vendors are the ones feeling it first. IBM won't be the last company explaining a bad quarter this way.

Chapter 4

Nova: The Verge AI got hands-on time with Samsung's upcoming smart glasses — the first real look at the hardware. Two new designs, developed with Google on the software side and, crucially, with Gentle Monster and Warby Parker on the design side. First specs include a 9-hour battery life. This is the most complete reveal Samsung has done for a wearable AI product.

Ray: Hardware reveals without pricing, without a software walkthrough, and without anyone actually wearing them in the real world for a week — that's the pattern of every failed smart glasses launch going back to Google Glass. A pretty render and a spec sheet is not a product.

Nova: Except the two things that killed every previous smart glasses attempt were battery life and social acceptability. Nine hours solves endurance — that's all-day wear. And bringing in Gentle Monster and Warby Parker isn't a cosmetic choice, it's Samsung saying: we know people won't wear tech that looks like tech.

Ray: The battery point lands. Google Glass died partly because you had to charge it mid-afternoon. And the fashion-brand partnerships are a genuine strategic departure — those aren't just licensing deals, they're Samsung admitting that eyewear is a fashion category first and a tech category second.

Nova: That's exactly the lesson Meta took with Ray-Ban Stories. The frames people actually want to wear are the ones that don't scream 'prototype.'

Ray: I'll say it directly: I came into this story treating it as standard hardware hype — no pricing, no real software detail, no real-world comfort data, same as every failed wearable launch before it. I was wrong to stop there. The 9-hour battery and the deliberate Gentle Monster and Warby Parker partnerships specifically address the two things that have sunk every prior attempt. I now think this is the most credible consumer wearable AI push to date. That said, credible reveal and shipping product are still two different things — final judgment waits for launch.

Nova: And the stakes are real. If Samsung ships this and it works, wearable AI computing stops being a niche and becomes the next platform. Google's involved, the eyewear brands are real, the battery is there. The window is open.

Chapter 5

Ray: Wired AI reports there's an active internal debate in the Trump administration over how to handle increasingly powerful Chinese AI models. No policy has landed yet — just debate. Which raises the question: is internal disagreement a sign that serious thinking is happening, or a sign that nobody knows what lever to pull?

Nova: Both, probably. The fact that the debate is happening at all means the administration recognizes Chinese AI capability is a real policy problem — not just a talking point. That's progress from a year ago when the conversation was mostly about chips and export controls.

Ray: Except the gap between recognizing a problem and having a coherent framework to address it is exactly where U.S. AI governance keeps stalling. Export controls on chips were the easy, tangible lever. What's the equivalent for AI models that can be downloaded, fine-tuned, and deployed without any hardware chokepoint?

Nova: That's the thread running through today's whole episode, actually. AMD and Anthropic are moving fast. Samsung is moving fast. And the institutions — the White House, regulators, standards bodies — are still in the debate phase. Capability keeps lapping governance.

Chapter 6

Nova: My takeaway: the Samsung glasses reveal is the signal to watch. When a chip giant, a software giant, and two fashion houses align on the same wearable product, that's not a prototype — that's a platform bet, and the next twelve months will show whether wearable AI finally has its iPhone moment.

Ray: Mine: IBM's mainframe quarter is the canary. Enterprise AI spending is already cannibalizing legacy hardware budgets, and every vendor still selling on-prem infrastructure should be asking whether their next renewal cycle survives the reallocation.

Nova: The open question — and it's a concrete one: if Samsung's glasses ship and sell, which platform owns the AI layer that runs on your face? Google's already in the partnership. Does that mean the wearable AI future runs on Gemini by default — and what does that mean for every other AI model trying to reach users?