Friday’s news that Meta Platforms is contracting for almost seven billion watts of future energy generation with three nuclear firms seems a watershed moment for the group.

Shares of the two public nuke companies partnering, Oklo and Vistra, closed up eight and ten percent, respectively. Meta shares rose a point on the news. Meta already had a deal with Constellation Energy, and Constellation shares closed up six percent Friday. The fourth supplier is TerraPower, privately held.

Also surging on Friday was the stock of the most mature nuclear vendor, NuScale Power, closing up four percent. Interestingly, Oklo’s fellow newbie, Nano Nuclear, closed down three percent.

For Meta, it’s about “de-risking” its ballooning appetite for power for its data centers, writes Rosenblatt’s Barton Crockett. The deals announced today, plus the deal with Constellation previously announced, add up to 7.7 gigawatts of commitments. Those deals “vault Meta to the top of the leader board in hyperscalers going nuclear,” writes Crockett.

Cloud companies “going nuclear” has an ominous ring to it, I must say.

Crockett lays out quite nicely the escalating power needs Meta has outlined, rising to as much as twenty gigawatts annually in a decade from now:

The artificial intelligence trade always loves more candidates for AI momentum, and a big group of these were the data center names that transitioned last year from being crypto-currency mining outfits to being AI data centers.

I noted in June, that larger outfits such as CoreWeave were trying to meet raging demand for AI hosting by subcontracting to firms such as Applied Digital, a trend that continued to rise throughout the summer, so that by October, analysts were cheering the rising prospects of Applied. 

Applied’s earnings report Wednesday evening, after market close, is adding to the fervor, driving shares up eight percent Thursday, to $31.93, giving it a return of two hundred and sixty-seven percent in the past twelve months. 

This is the third time in a row that the shares have surged on earnings, and the proximate cause is not simply the results but the enormous backlog of deals for data center capacity the company is bringing in.

I wrote at the beginning of last year that the market for “plain-old chips,” semiconductors that go into the diverse markets of consumer electronics, automobiles, industrial equipment, etc., was headed for a rebound. That argument was premised on the fact the vendors were clearing their books of a pile-up of inventory, and were seeing demand broadly recover in markets that had been weak for two years, including consumer electronics products. 

Turns out, I was a year early. 

2025 had mixed results for those vendors, which include Microchip, Texas Instruments, and Analog Devices, the last of which is one of the TL20 group of stocks to consider. The tariff and trade drama of 2025 added extra complications to what was going to be a cyclical rebound in the traditional analog and mixed-signal chip market. 

As a result, Microchip and TI trailed the broader market in 2025, while Analog Devices did just a little better, rising twenty-eight percent versus the Nasdaq’s twenty-one percent gain.

But Monday evening, after market close, brought the long-awaited good news that the plain-old stuff is now on that cyclical comeback. Microchip preannounced that its revenue for the December quarter will be “well above” what it had said back in November

The first week of the year begins with the Street trying to reconcile its excitement about the artificial intelligence spending boom with its anxiety about the potential implosion of that boom.

Monday, Morgan Stanley’s Lisa Shalett, writing for the bank’s Global Investment Committee, concludes that this year, it’s all about whether or not Artificial Intelligence will deliver a productivity payoff.

“All the bullish narratives of 2026 embed a singular known unknown—US corporations’ ability to drive meaningful productivity gains from GenAI,” writes Shalett.

She sees “apparent reasons to remain constructive” that AI will deliver such gains. The reasons include “stories” that Gen AI is “augmenting employee efforts,” stories that “are plausible and captivating,” writes Shalett.

Aggressive small-cap tech investors love to bet on long shots, even though returns can be hard to predict. Take the market for quantum computing, where returns are uneven and can seem somewhat illogical. 

IonQ, the company with the most revenue in the industry, rose only seven percent last year, trailing the Standard & Poor’s 500, while smaller competitors D-Wave and Rigetti rose two hundred eleven percent and forty-five percent, respectively. 

I don’t have much faith in these young companies given my skepticism about “scaling” quantum, as I explained in October. But, I’m just as fascinated with long shots in tech as anyone else.

One of my favorite long-shots in tech is not quantum but DNA computing, doing calculations with deoxyribonucleic acid, the stuff that makes up our chromosomes. I love the idea that we could compute via organic materials rather than silicon. The prospect of computing with our own human DNA, even inside of our bodies, is absolutely fascinating. 

I tend not to look too much at the life sciences industry because I don’t have the background to properly assess the technologies and the companies involved.

However, I’ve noticed one characteristic that fascinates me, which is the dependence of biotech companies on one of my favorite companies, Nvidia.

Pretty much everyone in the world is dependent on Nvidia at the moment, of course, and the escalating cost to companies such as Snowflake, discussed in the previous post, is obvious.

Less obvious is the cost to life sciences companies such as Oxford Nanopore, a maker of genetic sequencing machines. These are used in labs to detect the pattern of nucleic acids that make up the code of DNA and RNA.

I thought I’d offer up a few derivatives to the positive outlook for memory-chip investments and storage investments mentioned in the prior post.

To the memory and storage theme, I’d add companies with important products for managing “big data” that will be worth keeping an eye on. 

That includes TL20group of companies to consider member Snowflake, whose programs will be important for the loading and managing of vast amounts of data for AI-driven analytics. 

After a healthy forty-two-percent return in 2025, what is exciting for Snowflake in 2026 is that the company will be moving into the realm of “online transaction processing,” or, OLTP.

I expect we’re in for another big year for storage and memory technologies, which should be good for Micron Technology, Seagate, and Western Digital. 

That would be quite an achievement, given the fabulous returns on all of them in 2025, as listed in the table below. 

For DRAM memory, the plans of tech giants such as Google to spend billions on artificial intelligence infrastructure, including Nvidia chips and all the rest, is fueling a huge appetite for memory chips.

I mentioned a year and a half ago that technologies of memory are becoming increasingly important as memory becomes the bottleneck of artificial intelligence. The surging price of DRAM has borne that out, as per Micron’s quarterly report last month. 

Now, this is what I like to see at the start of a new year: the TL20 group of stocks to consider rising by a point on their first day while the major indices decline slightly. 

Among gainers Friday is TL20 name ASML, the maker of photo-lithographic equipment for chip produciton. ASML’s stock got an upgrade from Sell to Buy by Warren Lau of Aletheia Capital, after Lau raised his estimates for this year and next, expecting stronger demand for “extreme ultra-violet” lithography, or, EUV, the most advanced of ASML’s tools. Lau expects demand among DRAM makers, in particular, to be a driver of sales. 

This is a refutation of the “peak litho” story that was heard in June of last year, in which companies would buy less EUV equipment because they would instead simply package together simpler chips.

An interesting question for investors in Mega Cap tech is which of the three horsemen of cloud and artificial intelligence, Alphabet’s Google, Amazon, or Microsoft, will win out in the soaring market for AI in the cloud. The answer could be a big factor in their relative stock outperformance given that AI is going to be a big determinant of revenue growth but also expenses.

They all have very different business profiles, but I’m tempted to give Google the better odds here, with Amazon a distant second and Microsoft the least attractive of the three.

All three monetize AI in various ways, including the cloud computing services they rent, plus Amazon’s retail business and Google’s search and YouTube businesses. On the surface, it appears Microsoft has the best monetization strategy, as it now has ninety percent of the Fortune 500 paying money to use its “Co-pilot” AI assistant. Co-pilot allows Microsoft to charge more for each user of various software offerings including the “M365” productivity suite of Word and Excel, what used to be known as “Office.”

It seems everyone spent their Christmas holiday in the U.S. analyzing the deal by Nvidia to license technology from artificial intelligence startup Groq (not to be confused with the A.I. program Grok from Elon Musk’s xAI.)

As related by Bloomberg’s Ian King, Groq started life in 2016 as a chip maker, one of the many hopefuls convinced they could take share from Nvidia’s GPUs. After receiving $750 million in venture money, the latest bit in September, Groq is now licensing its chip technology to Nvidia, and proceeding in a different form as an independent company, running a cloud-based A.I. service for a fee.

Financial terms have not been disclosed, but the speculation is that Nvidia is spending as much as twenty billion dollars on this, which is not only a very large amount of money — Groq was valued at $7 billion following the most recent funding round — but also surprising given that Nvidia is getting a non-exclusive license to the technology.

I think the deal is not very significant, as I’ll explain in a moment, but the Street is generally enamored of the strategic virtues for Nvidia.

Vivek Arya was a product manager at the famed Bell Labs during the DotCom bubble in the late 1990s and early Naughts. He used to sell fiber-optic equipment to the likes of Worldcom and Level3. After the DotCom bust, he left for Merrill Lynch where he has been covering semiconductors for twenty-two years.

Friday, as part of Merrill Lynch’s year outlook, Arya made the case to myself and a small group of reporters why the current surge in artificial intelligence investing is different from the DotCom era.

His basic premise can be explained in a few simple observations: Cloud giants such as Google have to spend to not fall behind, and the AI stuff they’re building is not like the “dark fiber” at the heart of the DotCom bubble, a veritable field of dreams, it’s stuff they actually need right now because Google and cloud peers are “constantly capacity constrained,” and they can fund it with free cash flow …

We now come to the full story of profit as the new artificial intelligence concern. We have heard from Dell Technologies, Oracle, and Broadcom, and seen how each of them is dealing with pressure on their profitability as a result of the enormous cost of competing in the AI market.

And now we’ve had Micron Technology this week, which is conversely seeing its profit margin rise as a result of the surging price of DRAM and NAND chips it sells.

Shares of DRAM and NAND chip maker Micron Technology are surging by twelve percent Thursday morning to $253.04 after the company turned in Wednesday evening a first fiscal quarter report that the assembled analysts are calling “phenomenal.”

Astounding for the fact that memory chips are a commodity industry, Micron is now one of the only technology companies chasing artificial intelligence whose profit margin is going up at the moment, not down. And margins look set to keep going higher. That’s as tech giants from Broadcom to Oracle to Nvidia to Dell are seeing their profit margins shrink.

This, I think, marks a new and special high for Micron, not just a healthy chip cycle but a new period of increased profitability for the company, and possibly higher stock valuation.

Is the universe a hologram? Are our notions of “space” and “time” fundamental or emergent properties of the substance of reality?

Such questions are deeply pondered by quantum physicists. Fortunately, we investors have a simpler problem: valuing the three main quantum computing startups, D-Wave, IonQ, and Rigetti.

Such is the task set about on Tuesday by Jefferies & Co.’s Kevin Garrigan, who initiates the trio with Buy ratings for D-Wave and IonQ and a Hold rating for Rigetti.

“A rising tide lifts all boats,” Garrigan declares, writing that “in the current, early stage of quantum computing, ecosystem catalysts tend to lift sentiment and usage across architectures.”

Through seventy-one pages, Garrigan diligently lays out the shape of that industry in its early phases, with high rates of growth but off of a very small base of revenue.

DRAM and NAND chip maker Micron Technology is set to report fiscal first-quarter earnings Wednesday evening after the closing bell, and estimates and price targets have been on the rise for weeks.

The stock has been on a tear as a result, at a recent $242.36, up fifty-three percent in the past three months, and more than doubling from the reference date of the sixth rebalancing of the TL20 group of stocks to consider on August 29th.

The average estimate for the November-ending quarter is for $12.9 billion in revenue and $4 per share in net income. Those figures have risen from $11.7 billion and $2.97 at the end of August, so, a material increase.

And the change is even more dramatic in the estimates for the current, February-ending quarter: from $14.3 billion for revenue from $11.9 three months ago, and to $4.75 per share from $3.06 a share a month ago.

We saw earlier this week that surging artificial intelligence contracts for Oracle had a surprisingly negative effect on its stock as people doubted Oracle can profitably fulfill all those contracts.

Something similar was in the air on Friday for chip maker Broadcom, and it pushed the stock down by eleven percent to $359.93.

The big takeaway is that AI is putting pressure on everyone’s profit, an issue that’s going to continue to dog Broadcom and Oracle and others for the foreseeable future.

In brief, Broadcom CEO Hock Tan said, during Thursday evening’s fourth-quarter earnings call, that Broadcom, which produces custom AI processors, referred to as “XPUs,” for Google, Meta Platforms, and ByteDance (owner of TikTok), has won a fourth customer, AI startup Anthropic. That should be cause for celebration, except that this deal has different economics than the other three.

As I wrote three months ago, the most important thing for Adobe at the moment is that its arch-nemesis, recently public Figma, has been losing ground among investors, thereby opening a small window of opportunity for Adobe to regain the Street’s attention.

The drama of Adobe for a while now has been the worry that the company will see its franchise eroded as more and more average users, people not trained on its Photoshop and Premiere tools, turn to generative artificial intelligence to do their design work.

Figma has been the face of that threat, as it is young and “AI-first,” if you will, with creative tools built around Gen AI.

But investor interest in Figma has died out even as Adobe stock has stopped dropping.

Adobe stock closed up two percent Thursday after reporting quarterly results Wednesday evening. And since the September 11th report, the stock is up slightly at Thursday’s close of $350.43. (It is still down twenty percent for the year.)

Wednesday evening’s report by Oracle was fairly similar to the report three months ago: surging amounts of contracted revenue for cloud computing and artificial intelligence, and pressure on its profit margin.

This time, rather than a thirty-percent jump up in Oracle stock, the shares are down thirteen percent Thursday, at $191.97.

For many years, growth was the issue that dogged Oracle; now, with growth suddenly materializing, profit has become its bugbear.

As demand surges for Oracle’s offerings, pressure is only increasing for Oracle to prove that AI can be a profitable business.

Operating profit rose only nine percent, leaving aside the effect of currency fluctuations, on a thirteen percent increase in revenue, proof that spending to build Oracle’s cloud is coming in advance of revenue gains.

Adding to the pain, the company is going to spend fifteen billion more on capital spending this year than originally expected, for a total of fifty billion dollars, while making only as much revenue as originally promised. Lots of spending in advance of a payoff, in other words. And what Oracle has to deliver keeps rising: its backlog, “remaining performance obligation,” or, RPO, rose by sixty-eight billion dollars, fifteen percent, to $523 billion.

I hadn’t paid much attention to three-billion-dollar expense-management software maker Navan when it came public on October 30th in a billion-dollar offering lead by Goldman Sachs and Citigroup — which is just as well, as the stock has lost a third of its value from its first-day close of $20, and is down forty-five percent from the offer price, at a recent $13.85.

The company heads for its first-ever quarterly report on Monday, December 15th, and the Street is uniformly bullish on the software maker.

Navan mainly makes revenue from commissions paid to it by travel bookers when it directs business travelers to the bookers. There’s also an individual “per-transaction” fee direct from some customers, and a subscription fee for companies that use Navan’s expense-management software.

Some of this touches on something I didn’t know had a name: “Bleisure,” which is defined in its IPO prospectus by Navan as the “category of the business travel market defined by personal travel booked around or in connection with business travel.”

I’ve often thought International Business Machines is where things go to die, and another thing is going to that graveyard Monday morning, Confluent, the eight-billion-dollar maker of software to connect real-time events, which is being bought by IBM for a thirty-four-percent premium, at $31 in cash.

Confluent’s software, called “Kafka,” is used as a kind of connective tissue to let applications respond in real-time to events, such as a person clicking on a Web page. It was the force behind the feature on LinkedIn that tells you how many profile views you have, dating from back when Confluent CEO and founder Jay Kreps first wrote the Kafka code as an employee of LinkedIn. (More in my original profile of Kreps.)

Confluent has had mixed business results, and its stock has badly lagged the market for years now. My understanding is that the fundamental problems include the complexity of setting up a Kafka system for most enterprise IT, and also the fact that the early adopters tended to simply use the free, open-source version of Kafka rather than pay Confluent a fee.

I dropped Confluent from the TL20 group of stocks to consider in the February, 2024 rebalancing. Since that time, The TL20 is up 126% while Confluent is down two percent prior to this offer. Even with the thirty-four-percent premium, that would seem to make the decision the right one. You would have had to wait a long time to get bought out if you had stuck with Confluent.

“When you have a monopoly, you are in your own world […] And here, we created a very sophisticated system that now worries the Goliath.”

Software maker Snowflake’s drop of eleven percent on Thursday, to $235, following a decent fiscal third-quarter earnings report on Wednesday evening is a little like the twenty-six-percent plunge that Pure Storage took on Wednesday, which is to say, in my mind, it’s mostly about people taking profits.

The stock had been up seventy-six percent for the year heading into the report, after four quarters in which the stock jumped following each report, and a particularly strong second-quarter report that drove shares up back in August.

The proximate cause of the decline is that the company’s forecast was mixed. The projection for twenty-eight percent growth in Snowflake’s “product” revenue this quarter (the company never forecasts total revenue, as it leaves aside its services revenue) was higher than expected, but not as high as the “whisper” number. And the operating profit margin forecast of seven percent this quarter was below consensus for eight percent or higher.

Pure Storage is a stock that alternates between pretty stark increases and declines in stock price following earnings reports. Back in August, it saw a thirty-two-percent pop following its report. Wednesday, it’s a twenty-six percent plunge, dropping to $69.29.

The issue for Pure every single quarter is not the numbers, which continue to be very strong, it’s how much the company is going to sell to “hyperscale” clients, the cloud computing titans. So far, though CEO Charlie Giancarlo has repeatedly teased the prospect of deals for years now, Pure has only announced projects with Meta Platforms.

The sell-off today is that Giancarlo teased that he’s entertaining a variety of possible deal structures to lure more cloud clients, and one prospect is that Giancarlo will absorb up-front cost of NAND flash chips, the main ingredient, which could lower his gross profit margin — he’d take less profitable deals to win business, in other words.

“Our checks indicate that the new model can include PSTG procuring the NAND Flash directly from suppliers (Kioxia/MU) instead of only providing the software license,” as it does with Meta, writes Krish Sankar of TD Cowen.

We are still in the throes of a debate over an AI bubble. Skeptics point to the “circular” financing deals, optimists point to the raging demand in Nvidia’s most recent report.

Over the weekend, The Financial Times continued to dig into the details, with a story by Tabby Kinder and George Hammond pointing out that OpenAI has used its partners’ balance sheets to absorb some of the debt financing instead of OpenAI taking on that debt themselves.

Then, on Monday, another piece by Kinder and colleagues points out that OpenAI has taken a stake in private equity firm Thrive Holdings that is a non-cash swap: OpenAI will give Thrive’s portfolio companies access to OpenAI technologies, yet another instance of a circular deal.

I noted Monday the skeptical views about CoreWeave, the debt-laden AI hosting firm. MoffettNathanson warns of the risk to the company’s borrowing of having sub-prime customers such as OpenAI and Poolside AI, the French AI startup. They could drive up CoreWeave’s borrowing costs.

“Consumer demand of interactions with their brands, overall, is growing at double digits.”

Monday was the day for the analysts to weigh in for the first time on six-billion-dollar BETA Technologies, which came public on November 4th and has since lost a quarter of its value, dropping from its first-day close of $35 to $26.43, which is also below the IPO offer price of $34.

This is, I have to say, one of the more interesting IPOs in a while because they actually make stuff: Electric airplanes. The prospectus for IPO even looks cool because it has lots of pictures of electric planes.

The company raised a pretty good amount, $960 million, in the deal lead by Morgan Stanley and Goldman Sachs, and the coverage from the underwriters is almost uniformly positive today.

Eight-year-old BETA, which is based in South Burlington, Vermont, boasts that its planes, which have flown thousands of miles, including passenger flights into New York’s JFK airport, “represent a significant cost efficiency advantage,” the prospectus states. The “total operating costs are 42% lower compared to new traditional conventional aircraft based on internal estimates,” the company states.

Tuesday evening was a pretty terrible one for software, with shares of Nutanix, PagerDuty, Workday, and Zscaler selling off on mixed results. The stocks continued to fall hard during Wednesday’s regular session.

Dell Technologies was the saving grace Tuesday evening as its report showed some glimmers of hope in the company’s effort to make a profit off of AI servers. That was enough to drive the stock up six percent Wednesday.

A STRUCTURAL CHANGE FOR NUTANIX

Let’s start with the most disappointing report, Nutanix, which has been one of the TL20 stocks to consider since the September of 2023 rebalancing. The stock had been weak already this year, down four percent heading into Tuesday’s report. It closed down Wednesday by eighteen percent, at $48.34.

The proximate cause for the sell-off was the company missing with its revenue for its fiscal first-quarter ended in September, and cutting its full-year outlook for revenue.

Among Tuesday’s earnings winners is TL20 stock to consider Analog Devices, which beat expectations with both reported results and its outlook, sending shares up four percent at $249.05.

Among other winners Tuesday is Symbotic, the maker of robotics systems for warehouse automation, which is up thirty-six percent at $75.50. Symbotic had already been up a hundred and thirty-four percent heading into the report, so this is adding to massive gains already.

Symbotic is the result of a $4.5 billion takeover in 2022 of what had been Warehouse Technologies LLC., a private startup partly owned by Walmart, which was taken public in a reverse merger with one of SoftBank Group’s “special-purpose acquisition companies,” or, SPAC, a “blank check company,” back when SPACs were very, very active.

Symbotic tends to rise every other report, meaning, it flip-flops between big losses and big gains. In August it plummeted fourteen percent, today, it’s up much more than that.