
What happened
Barron's reported September 10 that Kepler Computing emerged from stealth with a memory architecture using ferroelectric materials and 3D stacking, claiming it can outperform conventional HBM using existing manufacturing tools. Intel is among its backers; samples are planned for later in 2026 and production in 2027.
Why it matters
Micron's bull case rests on scarcity — it already has HBM4 in high-volume production for Nvidia's Vera Rubin platform, and Reuters reported in June it had $22 billion of commitments from 16 strategic customers with supply constraints expected beyond 2027. Kepler represents the bear case that new architectures could eventually normalize those premiums.
What to watch
Kepler is not shipping commercial volumes today, so the threat hinges on whether it reaches production in 2027 — while Micron's existing HBM4 position and long-term commitments remain the near-term reality. Intel's exposure is strategic optionality, not an earnings engine, and short interest stood at roughly 2.6% of its float as of August 14.
WHO IT HITSMemory and semiconductor investors weighing Micron's scarcity-driven position against the possibility of competing architectures, and Intel shareholders assessing whether its Kepler backing translates into revenue or remains strategic optionality.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
Micron has become one of the AI boom's biggest beneficiaries because high-bandwidth memory has turned into a scarce strategic resource. Kepler Computing's emergence from stealth, reported by Barron's on September 10, directly challenges that scarcity by proposing a memory architecture built on ferroelectric materials and 3D stacking that it says can outperform conventional HBM using existing semiconductor manufacturing tools — a claim that, if borne out, could loosen the grip that incumbent HBM suppliers hold on AI system design.
The threat should not be exaggerated yet. Kepler is not shipping commercial volumes today, and Micron already has HBM4 in high-volume production for Nvidia's Vera Rubin platform, with its 36GB HBM4 product offering more than 2.8 TB/s of bandwidth and better power efficiency than its previous HBM3E generation. Reuters reported in June that Micron had $22 billion of commitments from 16 strategic customers and expected supply constraints to persist beyond 2027, giving it a substantial near-term cushion. The bear case is precisely what Kepler represents: extraordinary industry profits attract new architectures, capacity and competitors, and if alternatives eventually reduce dependence on conventional HBM, today's scarcity premiums could normalize rapidly.
For Intel, the connection is more speculative. A successful Kepler architecture could give Intel an indirect stake in a disruptive memory technology while complementing its foundry and packaging ambitions, but that investment does not make Kepler's future economics equivalent to Intel operating revenue. Hedge funds became more bullish on both stocks in Q2 — Micron ownership rose to 184 funds from 154 and Intel to 138 from 112 — while short sellers were slightly more cautious on Intel, with roughly 2.6% of its float short as of August 14 compared with about 2.7% for Micron. The real test is whether Kepler reaches production in 2027 and whether its claims hold up outside the lab; until then, Micron's production position and Intel's optionality remain the facts investors can act on.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Much of the attention on AI infrastructure buildouts is now tied to sheer compute power, with dominance define…

Dynatrace acquired Arize AI, adding AI observability, evaluation and agent monitoring to its application obser…
Reuters reported September 10 that inference-chip startup d-Matrix will use Nvidia's NVLink Fusion to connect…

Amazon announced Shop the Scene, which lets U.S

Mecka AI, which collects human motion data to train robots, is nearing a round led by Sequoia Capital at a val…

Lam Research said the semiconductor equipment market is "fundamentally sold out" and limited by clean-room cap…
