- Product & Design Pulse
- Posts
- Product & Design Pulse v80
Product & Design Pulse v80
AI's Gravitational Pull Hits Everything 🌀
Malcom Jones
February 23, 2026
Welcome to this week’s edition of Product & Design Pulse, where we explore the latest in tech, product, design, and innovation! Last week was about gravity — the kind AI exerts on everything around it. The global memory shortage dubbed "RAMageddon" made the physics visceral: hyperscalers are absorbing so much DRAM and NAND for AI infrastructure that consumer device prices are spiking, product launches are being delayed, and Micron has abandoned its consumer RAM business entirely — a zero-sum reallocation that Ben Thompson argues is accelerating a structural shift back to thin-client computing, where your local device matters less and cloud inference matters more. Meanwhile, Meta and NVIDIA formalized a sweeping multiyear infrastructure pact spanning millions of GPUs and next-generation CPUs, the kind of co-design lock-in that makes it increasingly difficult for anyone outside the hyperscaler tier to compete. On the software side, the week's signals were just as telling: AWS's own AI coding tool autonomously deleted a production environment and Amazon blamed the humans, while OpenAI poached Instagram's top celebrity liaison to begin repairing its fractured relationship with the creative industries whose work powers its models. The pattern is clear — AI isn't just reshaping products and workflows, it's restructuring the physical supply chains, infrastructure partnerships, and cultural relationships that the entire tech ecosystem depends on.
🎧 Audio Overview [BETA]
For those who don’t have time to read 😁 |
Last week…
AI's Hunger for Memory Is Starving the Rest of Tech
The global memory shortage — now widely dubbed "RAMageddon" — is accelerating as AI infrastructure buildouts from hyperscalers consume a disproportionate share of DRAM and NAND production, with prices spiking as much as 172% in 2025 and major OEMs warning of 15–30% device price hikes in 2026. The crisis is a zero-sum reallocation: every wafer directed to high-bandwidth memory for NVIDIA GPUs is one denied to consumer laptops, phones, and consoles, with Sony reportedly delaying the next PlayStation to 2028 or 2029 and Micron exiting its consumer RAM business entirely. For product leaders, this isn't a temporary supply hiccup — it's a structural repricing of the entire hardware stack, driven by three companies controlling 93% of the DRAM market and all three choosing AI margins over consumer volume.
Amazon's Own AI Coding Tool Took Down AWS — Then Amazon Blamed the Humans
AWS suffered at least two outages in December after its Kiro AI coding assistant autonomously decided to delete and recreate a production environment while attempting to fix a minor bug in AWS Cost Explorer. Amazon characterized the incidents as "user access control issues, not AI autonomy issues," even as employees internally warned that the company's high-speed AI rollout would cause "staggering damage." The episode crystallizes the emerging risk of agentic AI in production infrastructure — and the corporate instinct to protect the AI narrative by shifting blame to the humans nominally supervising it.
OpenAI Poaches Instagram's Celebrity Whisperer to Charm a Skeptical Hollywood
OpenAI hired Charles Porch, Instagram's VP of global partnerships for 15 years, as its first VP of global creative partnerships — a role focused on rebuilding trust with entertainment, music, and the creator economy. Porch, who orchestrated moments like Beyoncé's exclusive Instagram album launch and onboarding Pope Francis to the platform, will report to OpenAI's applications CEO Fidji Simo and plans to begin with a "listening tour" across creative industries. The hire signals that OpenAI recognizes its biggest distribution challenge isn't technical but relational — winning over the very communities whose work trained its models and whose goodwill it needs to monetize them.
Meta and NVIDIA Lock In a Multiyear, Multi-Million GPU Infrastructure Pact
Meta and NVIDIA announced a sweeping multiyear partnership spanning millions of Blackwell and next-generation Rubin GPUs, NVIDIA Grace and Vera CPUs, and Spectrum-X Ethernet networking across Meta's hyperscale data centers. The deal also includes NVIDIA Confidential Computing for WhatsApp's AI features, enabling on-device-level privacy guarantees in a cloud inference environment. For the broader market, the partnership reinforces that AI infrastructure dominance is being locked in through long-term co-design agreements — making it increasingly difficult for smaller players or alternative chip architectures to compete for hyperscaler workloads.
The AI Era Is Reviving the Thin Client — and Making Your Devices Less Important
Ben Thompson argues that AI is fundamentally reversing decades of thick-client dominance: chat interfaces and agentic workflows require almost no local compute, shifting the locus of value back to centralized data centers with superior models, larger context windows, and faster inference. The global memory shortage compounds this dynamic — as AI infrastructure absorbs available DRAM and NAND supply, consumer devices become simultaneously more expensive and less essential, with Sony potentially delaying the PS6 and PC prices rising across the board. The strategic implication is that the companies controlling cloud inference capacity and the memory supply chain powering it are building structural advantages that may outlast the shortage itself, as path dependency locks new workflows into thin-client architectures before local inference becomes viable.
