Hurtling towards the holidays, but I’ll be working this week and looking forward to earnings from Jabil (JBL: NYSE), reporting after dumping both their CEO and the low-margin Apple business and Micron (MU:NYSE), with its AI-focused High Bandwidth Memory taking DRAM market share.

Our reports last week on Mongodb (MDB: NYSE), Oracle (ORCL: NYSE), C3.ai (AI: NYSE), Adobe (ADBE: Nasdaq), 

Broadcom (AVGO: Nasdaq)  are available to clients, with video summaries on YouTube and, of course our popular summaries of the summaries on Instagram and Tiktok. 

(A calendar of our upcoming video reports at the end of this note).

As always, I’m focused on three things:
1) Technology-driven change;
2) the latest in innovation and startup trends, and;
3) stock fraud.

In This Note:

• Micron Stacks the AI Deck

• Google’s Quantum Leap of Faith

• Epistrophy In The News

• Upcoming Earnings In Focus

• My Plans This Week

Micron Stacks the AI Deck

After years of playing catch-up in the high bandwidth memory market, Micron Technology (MU: NASDAQ) has accomplished something remarkable: they're sold out of HBM through 2024 and most of 2025. Supply constraints of this magnitude for a new memory technology stand unprecedented.

The reason? NVIDIA (NVDA: NASDAQ) has secured Micron's entire HBM3E production for its H200 AI accelerator. This validates Micron's technology investment while exposing a vulnerability in AI's supply chain.

We’ll likely hear more about that with this week’s results. Micron's HBM3E delivers 1.2 terabytes per second of bandwidth at lower power than competing 24GB solutions from Samsung Electronics and SK Hynix. Micron’s new 12-high stack configuration provides 36GB of capacity - a 50% increase crucial for training larger AI models.

This product is finally driving Micron’s margins, thanks to HBM's premium pricing compared to conventional DRAM.

Manufacturing creates bottlenecks. HBM requires advanced packaging technologies like chip-on-wafer-on-substrate (CoWoS) and precise integration of through-silicon vias (TSVs). Micron's HBM3E uses twice as many TSVs as current HBM3 products. This complexity means HBM production consumes approximately three times the wafer capacity as DDR5 memory for the same number of bits.

Technical challenges multiply. Micron has incorporated fully programmable memory built-in self-test (MBIST) capabilities that run at full specification speeds, enabling faster validation and better reliability. The precision required for stacking 12 dies with perfect alignment measures in microns, where a single misaligned connection in thousands can render an entire stack worthless.

But market dynamics in memory can shift rapidly. Samsung recently announced qualification of its HBM3E with a major customer, while SK Hynix maintains nearly 50% market share. Micron expects “several hundred million dollars” in HBM revenue for fiscal 2024, targeting 20-25% market share by 2025. Their advanced 1-beta process node provides a foundation for these goals, assuming yields remain stable.

Memory bandwidth constrains AI workloads. This drives Micron's development of HBM4 and HBM4E. Yet once Samsung and SK Hynix scale HBM3E production, traditional memory pricing cycles will likely emerge.

Micron has delivered a technical achievement in HBM3E. Whether they can maintain their advantage against intense competition remains uncertain. The battle for AI memory supremacy accelerates.

Google’s Quantum Leap of Faith

Alphabet’s Google (GOOG: Nasdaq) says its new "Willow" quantum processor performed a target computation in just 5 minutes—a task that would take the world's fastest classical supercomputer over 10 septillion years, exceeding the age of the known universe. 

While a remarkable scientific achievement, Willow's real-world impact remains uncertain. The benchmark used, random circuit sampling, has no practical applications. And though Willow cut error rates to an unprecedented 1 in 1,000 per cycle, experts estimate we'll need error rates of 1 in 1,000,000,000,000,000,000  (one quintillion) and 100-1,000 times more qubits to tackle useful problems. Willow has just 105 qubits.

And yet investors pour in. Alphabet’s value jumped $113 billion on this news! Quantum startups raised a record $1.5 billion in venture funding across 50 deals in 2024, according to Crunchbase, nearly doubling 2023's total and far surpassing the previous high of $963 million in 2022.

In January, Colorado-based Quantinuum raised $300 million at a $5 billion valuation. PsiQuantum, a Palo Alto startup, secured a $620 million package from Australian government entities in May. The U.K.'s Riverlane landed a $75 million Series C in August. 

Then there's IonQ (IONQ: NYSE), which went public in 2021 at a $2 billion valuation and now sports a hefty $7.4 billion market cap – despite booking a paltry $11.1 million in revenue for fiscal 2024. To be sure, the company scored a $13.4 million contract from the U.S. Air Force in December 2022, the largest ever federal award for quantum technology: that says something right there, that the largest contract ever is equal to 20 minutes of Alphabet revenues. 

Some of the buzz may be spillover from the red-hot AI sector. In theory, quantum computers could dramatically accelerate the development of artificial intelligence by performing certain computations much faster and more efficiently than classical machines. They might also slash the enormous energy consumption of today's data centers used for AI training.

But let’s be clear: this is a science fair project, a really expensive science fair project. These are expensive toys with few practical uses. Building a large-scale, fault-tolerant quantum machine will require overcoming enormous technical obstacles. Quantum bits, or qubits, are extremely fragile, losing coherence at the slightest disturbance. As systems scale up, error correction overhead balloons exponentially in complexity. 

The risks loom. A powerful quantum computer could pose major dangers. It might smash today's cryptography, exposing sensitive data and crippling digital security. It could disrupt cryptocurrency markets by accelerating the mining of Bitcoin and other tokens. Some worry it could even enable the creation of new kinds of weapons.

The flood of capital into quantum computing reflects the technology's tantalizing potential. But it also highlights a growing disconnect between hype and reality, as geopolitical FOMO and Silicon Valley's hunt for the next big thing drive eye-popping valuations for startups long on buzzwords but short on substance.

Google's Willow chip marks an exciting milestone on the long road to practical quantum computing. And the quantum revolution may indeed be coming—but not anytime soon.

Epistrophy In The News

My Oracle (ORCL: Nasdaq) preview on the Schwab Network turned out to be prescient. I talked about potential lumpiness in the quarter and yet strong growth building on the modular design  of Oracle’s data centers – exactly what the company cited on the call. 

And what fun talking about the quantum computing hype coming out of Alphabet with Julie Hyman and Josh Lipton on their terrific Yahoo! Finance show. 

Finally, about a year ago we decided to take our Drill Down Podcast and put it on video. I had great trepidation. My standards for TV production, having been coached by a handful of great camera ops and producers at CNBC and Bloomberg, are high. But we threw together some makeup, hair gel, jerry-rigged lights and of course my trusty RØDE mic, and pretty quickly saw exploding engagement on iTunes, on YouTube and lots on Spotify. 

But I had no idea how successful this was becoming. This last week we were informed that we’re now in the top 10% of Spotify video podcasts. 

Shocked and honored.

Upcoming Earnings In Focus

My Plans This Week

I'll be in San Francisco through the end of the year and quite available, though out of pocket this Wednesday.

I hope these thoughts are helpful to you. I’d love to discuss them further and, as always, comments, questions and ideas are appreciated.