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.
Clip via the episode · @DaveBlundin @positron_ai · @PeterDiamandis @moonshots_pod
The bottleneck shifted. Most people haven’t noticed.
Everyone is watching the GPU race — Nvidia, AMD, custom silicon. But according to Dave Blundin, that’s the wrong race to watch. His argument: the real choke point in the AI stack is now HBM (High Bandwidth Memory), the exotic, expensive memory that sits alongside inference chips.
Positron’s $5B valuation — built on chips that sidestep HBM entirely — is Blundin’s evidence. Valuations that size don’t happen on theory alone.
The Constraint Shift
If Blundin’s thesis holds, winning the GPU war while losing the memory war is still losing. The scarce resource in AI infrastructure isn’t raw compute — it’s the memory bandwidth to feed that compute.
Three words. The strategic implication is significant: optimizing for every constraint simultaneously is impossible. In Blundin’s framing — as we read it — the chip designers who tried to preserve every performance metric while dodging HBM probably failed. Positron apparently made deliberate trade-offs, accepted degraded performance on some dimensions, and unlocked a different cost and supply structure entirely.
That’s a classic constraint-led innovation move. When a scarce input (HBM) becomes the binding limit, the winning strategy isn’t to fight for more of it — it’s to redesign around it. The quote suggests Positron did exactly that, knowingly accepting engineering compromises to escape the HBM bottleneck.
Map of AI — Stack Position
In FourWeekMBA’s Map of AI framework, Positron sits at Layer 2: Infrastructure & Compute Enablement — the hardware layer that everything above it depends on. Shifts at this layer tend to ripple upward into model economics, inference pricing, and ultimately which AI products can profitably scale.
The Structural Read
If inference chips that avoid HBM can reach a $5B valuation today, the market is pricing in a future where HBM supply remains constrained and expensive. That’s a bet on structural scarcity — not a short-term supply chain hiccup. Watch whether other inference chip startups converge on the same HBM-avoidance thesis.
Positron — By The Numbers (per episode)
$5B
Reported Valuation
~$875M
Raised (approx.)
Figures cited by Dave Blundin in the episode. Not independently verified by FourWeekMBA.
The Bottom Line
Blundin’s argument — as expressed in the episode — is that the GPU narrative is a distraction. Memory bandwidth is the real gate. If he’s right, the companies willing to “accept some bad things” and design around HBM may own a structural cost advantage that GPU-optimized competitors can’t close. That’s worth watching closely.
Clip via the episode — 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 · @PeterDiamandis @moonshots_pod
This is editorial analysis of a public podcast clip, not investment advice. All figures cited are attributed to the speaker, not verified independently by FourWeekMBA.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
