Clip via the episode โ Dave Blundin’s thesis: the binding constraint on AI progress is now HBM memory, not GPUs. His proof: Positron hit a $5B valuation (~$875M raised) building inference chips that avoid HBM. / @DaveBlundin @positron_ai ยท @PeterDiamandis @moonshots_pod
Everyone is staring at the GPU scoreboard. Blundin’s argument โ expressed in that deceptively short phrase โ is that the real constraint has already moved one layer up the stack, to High-Bandwidth Memory.
That reframe, if correct, doesn’t just reshuffle chip rankings. It restructures where value accrues across the entire AI infrastructure layer.
The classic assumption: more GPUs = more AI progress. Blundin’s counter: once compute is abundant enough, memory bandwidth becomes the binding limit. You can’t feed a fast chip what it can’t retrieve fast enough.
“Accept Some Bad Things” โ three words that imply a design philosophy: sacrifice HBM dependency even at the cost of some capability, because the bottleneck you remove is worth more than the performance you trade away.
THE POSITRON SIGNAL
Blundin points to Positron โ inference chips architected to sidestep HBM โ reaching a ~$5B valuation on roughly $875M raised. That market verdict is the evidence he cites. Capital is betting the constraint is real and the workaround is viable.
POSITRON ยท AS CITED BY BLUNDIN
~$5B
Valuation (cited)
~$875M
Raised (cited)
FOURWEEKMBA ANALYTICAL READ
FDE Framework lens: Positron sits in the Enabler tier โ not building AI products, not distributing them, but removing the infrastructure friction that limits everyone else. Enablers that solve a binding constraint at scale tend to capture outsized margin before the layer commoditizes.
Map of AI lens: If the HBM layer is genuinely the new chokepoint, the infrastructure sublayer just bifurcated โ HBM-dependent vs. HBM-agnostic. That’s a structural fault line, not a product difference. Architects who design around the constraint early hold the durable position.
THE BOTTOM LINE
Blundin’s three-word phrase is a design mandate dressed up as a shrug. If the binding constraint really has shifted from compute to memory bandwidth, then the companies willing to accept the trade-offs of HBM-avoidance aren’t compromising โ they’re positioning. The market data he cites suggests that bet is already being made, at scale.
Clip via the episode โ Dave Blundin’s thesis: the binding constraint on AI progress is now HBM memory, not GPUs. His proof: Positron hit a $5B valuation (~$875M raised) building inference chips that avoid HBM. / @DaveBlundin @positron_ai ยท @PeterDiamandis @moonshots_pod
This is FourWeekMBA structural analysis of a public podcast clip โ not investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
