Taiwan Semi lowered its outlook for the chip industry this year, but its sales are still expected to rise on the AI chip race, Oracle may have to spend like crazy to stay in the AI game, and Zscaler’s competition from Microsoft and others looks not as bad as previously thought.

Shares of Netflix are down almost six percent in late trading despite the company beating on revenue and profit with its Q1 report, and forecasting this quarter’s profit higher as well, with new subscribers coming in well above consensus.

Seems that after a twenty-five percent run-up this year, people are taking profits on what looks to be less upside in the current quarter. Tonight’s drop is the reverse of the large gains the stock saw the prior two quarters. The giant addition of subscribers in the January report boosted the stock by eleven percent at the time, but, clearly, adding subscribers doesn’t always carry the day for the stock.

But there’s also a disheartening surprise in the forecast section of the shareholder letter: Netflix will stop disclosing the number of members next year.

ASML Holding, one of the top chip-equipment makers in the world, missed with its first-quarter revenue report Thursday morning, and missed with its outlook for this quarter, but the bulls on the stock are taking it in stride.

“We attribute the shortfall to quarter-to-quarter lumpiness, and do not see a demand issue,” writes Raymond James analyst Srini Pajjuri in a note to clients Thursday.

“A near-term hiccup” is how TD Cowen analyst Krish Sankar describes the report. The focus of the Street is not so much revenue as is it is the company’s number for “bookings,” the contract value signed during the quarter. Bookings are a prelude to revenue, as the product is delivered and the money is formally recognized on the income statement.

Those bookings, at €3.6 billion, were well below expectations for about five and a half billion euros.

• Earning season is getting underway as I record this, with the early reports IBM, Netflix and ASML.

• The week that ended April 12th saw some sell-offs in names that include Nvidia, one of the TL20 stocks worth considering. Nvidia ended the week of the 12th up fractionally, reversing a 3% loss the ending April 5th.

• I had written about Nvidia’s plight on April 4th. It's what I call “inverse estimates behavior.” The stock is way down from a recent high of about $950, and over the course of the past two weeks, it has really meandered through ups and downs.

• I can't tell you when it will relieve itself, but I will tell you that estimates have continued to rise and my inverse estimates behavior theory for Nvidia would be that it will see some gains when some of these increases slow down.

The people who sold the Reddit IPO don’t believe in its immediate future, Amazon is your best bet among social and commerce stocks this earnings season, and Enovix announces positive development amidst nail-biting.

Things keep getting worse for Tesla heading toward its earnings report, Cisco will be fine after the speed bump, the Street loves Astera Labs’s expensive stock, and Informatica and Salesforce are each other’s problems at the moment.

“We are at that kind of watershed moment in quantum where you transition from research experimentation to business benefit.”

The news flow of late is filled with big dollar signs for the potential investment in AI. They range from Advanced Micro Devices CEO Lisa Su’s comment in December that the market for AI chips could be worth four hundred billion dollars by 2027 to the report two weeks ago from The Information that OpenAI is contemplating building hundred-billion-dollar AI computers.

What to believe? The average individual will note that many AI programs have issues, such as “hallucinations.” Either the money is therefore good money after bad or it’s a desperate attempt to fix a big mess.

But there is another view to consider.

Tesla could transition away from the car business, Amkor is less expensive than a lot of AI stocks, there’s a reason ASML is pricier than many other chip stocks, and Meta’s trends are looking good now and through 2028.

We are just approaching the very beginning of earnings season — how time flies! — which gets underway a week from Thursday, when Taiwan Semiconductor Manufacturing and Netflix will both report results, the traditional early birds of the season.

And that means it’s time for earnings upgrade season, when analysts tune up their spreadsheets, and, if they’re feeling chipper, raise price targets and ratings on stocks.

Nvidia’s having another lousy week, falling about two percent so far, though it is recouping some losses Friday afternoon. That follows a four percent decline the week prior. Thanks to the sudden surge in the stock this year — still up almost eighty percent — the market is anxious about Nvidia being a “bubble,” part of a bigger bubble in all things artificial intelligence.

But there’s another factor to consider as well, slightly less emotional, slightly more mechanical. As I noted last fall, Nvidia tends to trade for a period of time opposite to how revenue and earnings estimates are running. When estimates go up, the stock sells off, and the opposite if estimates remain flat or go down, though Nvidia estimates don’t go down too much these days.

• The abbreviated holiday week that ended Thursday the 28th of March, with Friday closed for Good Friday for most markets, saw the Nasdaq Composite Index down slightly, about 0.3%, the Standard & Poor’s 500 Index managing to eke out a 0.3% gain, and the TL20 group of stocks worth considering, down 2.5% — weighed down mostly by Nvidia, which saw a decline of 4% for the week. It's one of the largest weekly declines for Nvidia in a while that I can remember.

• I mentioned in last week's podcast that a number of observers have been warning of frothiness in the AI market, and so there's been some pressure on Nvidia shares of late. I continue to think it is a very valuable company and a very valuable stock to own and that it is not overvalued by most measures.

• It was a mediocre week for Nvidia, but it was the prelude to a terrible week for Tesla. Sales have been seeking out of the EV market like a balloon losing air. First it was a warning from Aehr Test Systems, supplier to the chip makers for EVs, ON Semiconductor and Wolfspeed.

“I am not just selling something to secure some piece of technology, but I am trying to secure that entity […] That’s the kind of longevity and mindset that lets you get a company to a billion dollars and beyond.”

Intel on Tuesday evening offered up an eagerly-anticipated re-categorization of its income statement. There’s a huge operating loss in the middle of those numbers. The reality that it will take Intel years to close that gap has had the effect of a nearly seven-percent sell-off in the stock Wednesday morning.

Analysts are debating, as they often do, Is a huge operating loss a glass half empty or half full?

The new numbers reflect Intel’s decision, announced last summer, to report the financials for its factories, its “foundry,” as a stand-alone business — what I called Le Divorce. The point of Le Divorce is to say what belongs to whom — to make the foundry accept its own profit and loss, separate from chip product sales, as CEO Pat Gelsinger builds the foundry into a third-party contract house, “Intel Foundry Services,” serving not just Intel but also competitors.

Tesla showed its worst miss on deliveries in years, which is probably from much weaker demand in the U.S., solar energy equipment maker Array Technologies has good prospects for a comeback, and BE Semi and Alphawave Semi are best Ides in the AI chip race among European stocks.

Analysts are salivating over the opportunity for fiber-optic components in artificial intelligence, Semtech is one of the big beneficiaries, Micron Technology is going to see a big lift from “HBM” memory chips, and Zscaler and others are in the early innings of the “SASE” change in cyber-security.

Electric vehicles may have seen their first quarterly decline in sales in four years, another bad sign for Tesla, Palantir’s stock has grown too pricey on AI hype, the big annual fiber-optics show reflects a still-struggling industry, and chip maker Monolithic Power’s doing a lot better than rumor suggests.

The week that ended March 22nd was generally a favorable one for markets, with the Nasdaq Composite Index rising 3%, and the Standard & Poor’s 500 Index rising 2%.

The biggest news of the last two weeks was the amazing performance of Micron Technology. In the week ending the 22nd, Micron shares surged 18%. The stock since the start of the year is up 29%. The proximate cause was the company’s earnings report on March 20th, in which it said that demand is improving for both DRAM and NAND chips.

Micron CEO Sanjay Mehrotra said that prices for chips will be improving through the rest of this year and gross profit margin will be improving as a result of both those prices and because of the improving “mix” of chips — specifically, more ““high-bandwidth memory,” the stuff that’s needed for the fastest AI chips.

“We have a number of new growth vectors that will start contributing in fiscal year ’25, but it takes a while for them to make a meaningful impact.”

Microsoft’s Copilots may boost cloud revenue next year, Reddit’s stock price is too high even in a bullish scenario, hard disk drive maker Seagate has an AI angle, and how to “rent” AI stocks with options trades.

Back in the olden days, a social media startup had to have massive growth and be profitable. That was what Meta, then Facebook, offered public markets when it came public in 2012: annual revenue growth of eighty-eight percent, an operating profit margin of forty-seven percent, and an amassed equity position — the retained earnings of the business — of over five billion dollars, before any IPO proceeds.

As the shares of another social media startup, Reddit, soared in its stock market debut on Thursday by fifty percent, the common refrain was that the business is not profitable. That’s an understatement. Next to Reddit, Facebook’s IPO looks like a golden era. After almost as many years in business as Meta, Reddit’s business is lousy — strikingly lousy.

Reddit has neither profit, nor really impressive growth, and a huge amassed deficit. I’ve put the numbers side-by-side from Reddit’s prospectus this week with that of Meta back in 2012, with some annotations.

“Wireless technology has advanced to the state where, five years from now, we're going to look back and people aren't going to blink when they say whether it's wireless or fiber — it’s not going to matter.”

Broadcom is the industry’s alternative to Nvidia and AMD, DeFi is an unusual bet on the rise of Bitcoin, SoundHound’s “black box” business is not something you want to bet on for AI, Microsoft could benefit from Nvidia’s software innovations, and now is the best time for software investors since 2019.

Tuesday and Wednesday had seen a rush of positive notes about memory-chip maker Micron Technology, and the enthusiasm was finally vindicated Wednesday evening as the company offered an earnings report, and quarterly forecast, that positively crushed consensus.

Micron’s revenue of $5.82 billion was well above consensus of $5.34 billion, while EPS of positive forty-two cents topped the consensus for a loss of twenty-five cents. The outlook for this quarter, moreover, is for revenue in a range of $6.4 billion to $6.8 billion, well above consensus for $6 billion, and profit of 38 cents to 52 cents, versus consensus for 20 cents.

It was the fifth quarter in a row in which the company’s forecast topped expectations. Best of all for investors, Micron delivered a higher-than-expected gross profit margin of twenty percent, up from barely one percent the prior quarter, and negative thirty-one percent in the year-earlier quarter. Gross profit margin is closely watched as an indicator of the health of Micron.

A report Monday by Bloomberg’s Mark Gurman, citing unnamed sources, stated that Apple is in talks with Alphabet’s Google to bring the latter’s artificial intelligence mega-program, Gemini, to the iPhone. Gurman’s report raised but did not answer an interesting question, Who pays whom?

Gemini is a recently-released program from Google that can do things similar to OpenAI’s ChatGPT, but also a whole lot more, such as finding a particular moment in a video clips based just on the description the user types. The program can be used for free, and Google offers an higher-capability version, called “Advanced,” as part of a Google subscription that includes storage and Google Docs and other things, for $19.99 per month.

Gurman’s report stirred a number of enthusiastic reflections from the Street in the ensuing twenty-four hours. None of those reports is confident about the question of who’s the one paying here.

Nvidia’s CEO Jensen Huang’s new AI chip is boosting price targets, it’s going to take a while for Zoom stock to be appealing again, you should buy DevOps firms GitLab and JFrog as Gen AI spreads, Duolingo wouldn’t be killed by a TikTok ban, and Zscaler and CrowdStrike can benefit from the fallout of that massive data breach at Change Healthcare in February.

In recent years, subtly but surely, the path by which Nvidia has confounded observers has been not its strength in making gigantic artificial intelligence chips, which is formidable, but rather the software it uses to tie it all together, software that is now becoming a billion-dollar annual business.

An intriguing example came during an event in San Jose, California, on Monday. It was the first conference Nvidia has held since the COVID-19 lockdowns, “GTC,” an event that the Street, gushing, referred to as the “Woodstock of AI” or the “Lalapalooza of AI” (analysts come from varying generations.)

Co-founder and CEO Jensen Huang offered a shopping bag full of interesting technical unveilings. The main event was a new chip design, “Blackwell,” meant to run the largest neural networks of AI. It multiplies ten-fold the number of floating-point math operations per second compared to the predecessor “Hopper” chips. It will also be several times faster in that regard than the competing part on which rival Advanced Micro Devices has placed its hopes, the “MI300X.”