Tech Brew Ride Home - Fri. 08/02 – Intel Is Bailing Water

Episode Date: August 2, 2024

So, tech earnings continued to be boring. Apple was fine. Amazon was not quite so fine. And then someone we don’t cover closely really messed the bed. I’ll tell you who. Are we starting to see the first next-generation AI startups? I’ll tell you about an interesting new model. And, of course, the Weekend Longreads Suggestions. Links: Amazon Beats On Earnings, Revenue Comes In Light Despite Strong Cloud Performance (Investor's Business Daily) Intel is laying off over 15,000 employees and will stop ‘non-essential work’ (The Verge) 23andMe Board Committee ‘Disappointed’ in CEO Anne Wojcicki’s Take-Private Proposal (Bloomberg) Stable Diffusion creators launch Black Forest Labs, secure $31M for FLUX.1 AI image generator (VentureBeat) Weekend Longreads Suggestions: A $500 Open Source Tool Lets Anyone Hack Computer Chips With Lasers (Wired) The AI Keeps Score (The Verge) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech Meme right home for Friday, August 2nd, 2024. I'm Brian McCullough today. So tech earnings continued to be boring. Apple was fine. Amazon was not quite as fine. And then someone we don't cover closely really mess the bed. I'll tell you who that was. Are we starting to see the first next-gen-gen-gen-a-I startups? I'll tell you about an interesting new model. And of course, the weekend long-range suggestions. Here's what you missed today in the world of tech. I was going to say tech earnings season was back to being boring for us. but then somebody went and spoiled it. For example, starting with Apple, it was fine.

They bounced back a bit from the bad earnings of last quarter, beating expected results in just about every category. I won't go through all the numbers. They're basically what was expected, though iPad sales were up 24%. A new lineup can do that. And revenue from services, where the App Store, Apple TV Plus, and Apple Music All Live,

rose 14% to $24.21 billion, a new all-time record. But Amazon was the first to throw me a curveball. Amazon shares are down 12% as I write this, although the overall market is down quite a bit as well. Again, I don't know that going into the numbers is that important, but let's just say the street was not pleased by the forecasted Q3 revenue, which came in below estimates. Also, Amazon reported slowing online stores sales growth of just 5% year-over-year in Q2 versus a growth of 7% in Q1. quoting investors business daily. Amazon also projected lower operating income for the current quarter than Wall Street expected. The company guided for operating income of $13.25 billion at the midpoint of its range for the September quarter compared to analyst expectations of $15.3 billion. Still, the Tech Giant's second quarter report highlighted the company's progress in its push to boost profitability. CEO Andy Jassy has worked to cut costs during his three years leading the company.

Operating income increased 90% year-over-year to $14.7 billion. the company said, analysts were projecting operating income of $13.7 billion from Amazon prior to the report. AWS provided $9.3 billion of that operating income, but the $5.1 billion in operating income from Amazon's North American retail operations, which has been a focus of investors, came in below expectations of $5.4 billion, according to facts set. While Amazon Web Services revenue exceeded expectations, other segments of Amazon's business saw sales below expectations in the second quarter. Amazon's advertising business grew sales 20% to $12.8 billion. Analysts were forecasting $13 billion in sales for the division, according to fact-set. Advertising has emerged as Amazon's

fastest-growing division and a key source of retail profits, end quote. So I guess if online commerce is hitting tailwinds and your fastest-growing segment is not growing as fast, that will do it. But that's not even the biggest curveball of the day. This was, we don't always cover the them, but at the time of this writing, Intel stock is down 30%. You heard me. Intel reported that Q2 revenue was down 1% year over year to $12.8 billion versus $12.9 billion estimated data center and AI revenue was down 3% year over year to $3 billion, and the company gave Light Q3 guidance. Intel also announced plans to cut more than 15% of its workforce, the majority by the end of this year, as part of a $10 billion cost reduction plan and is suspending its dividend starting in Q4.

Yeah, doesn't sound so great over there, but also announcing huge layoffs usually makes the street happy. So what the heck is happening under the hood that is causing a 30% drop? A decline of this magnitude where it to hold through the regular session on Friday would be Intel's worst since the stock fell 22% on September 22,000. Well, I don't know. MIST mobile, they missed arm. Could it be that investors are realizing they are going to miss out completely on the AI Gold Rush, too? Quoting the verge. Intel will reduce its R&D and marketing spend by billions each year through 2026. It will reduce capital expenditures by more than 20% this year. It will restructure to, quote, stop non-essential work, and it'll review, quote, all active projects

and equipment to make sure it's not spending too much. This is painful news for me to share. I know it will be even more difficult for you to read. Reid's part of a memo from Intel CEO Pat Gelsinger to staff, which you can read in full at the bottom of this post. The company just reported a loss of $1.6 billion for Q2 20204, substantially more than the $437 million it lost last quarter. Our Q2 financial performance was disappointing, even as we hit key product and process technology milestones admitted Gelsinger in the company's press release.

Our revenues have not grown as expected, and we've yet to fully benefit from powerful trends like AI, he writes in his employee memo. Second quarter revenue was $12.8 billion down just 1% year over year, and it's not like all of Intel's businesses are failing, while Intel has absolutely been losing money on its chip-making foundry business as it invests in new factories and extreme ultra-violent lithography, to the tune of $7 billion in operating losses in 2023 and another $2.8 billion this quarter, the company's products themselves aren't unprofitable. Almost all the losses this quarter and last quarter came from Foundry, while its sales continue to stay relatively stable, and its PC and server businesses stay profitable. The company is also set to receive up to $8.5 billion in U.S. government funding from the Chips Act, but investors didn't seem happy that the company kept itself on a knife's edge.

Over the past two years, before this quarterly loss, it had continued to swing between losses and profits overall for just $1.1 billion in cumulative losses between Q2, 2022 and Q2, and Q2. 1-2020-24. Intel is now the worst-performing tech stock in the S&P 500 this year, CNBC wrote in April. From a tech leadership perspective, Intel's not yet a big player in AI server chips like Nvidia, maybe not even a notable small one like AMD. Its relatively recent entry into graphics has yet to impress, and it had to overhaul its flagship laptop chips significantly to address the existential threat of armed ships from the likes of Qualcomm and Apple, which can offer more battery life than Intel. Like competitors, the company now, partially relies on TSM, not just its own foundries, to help produce some of its most advanced

chips. Microsoft recently followed Apple's lead in ditching Intel chips for its latest slate of consumer hardware, including the Surface laptop and Surface Pro, and launched its co-pilot plus PC initiative exclusively with Qualcomm without waiting for Intel or AMD's new flagship laptop chips to join them. Intel is currently dealing with two generations of potentially defective desktop CPUs, though the company currently believes it can mitigate the issue with a software update and doesn't currently plan on recalls. On the company's earnings call today, Intel CFO David Zinsner, just suggested that the company's next flagship AI laptop chip, Lunar Lake, won't be enough by itself to turn things around. While he says, quote, the AIPC is a

big winner for the company and Intel plans to, quote, ramp that product significantly next year to meet market demand, he also described Lunar Lake as a, quote, narrow targeted product that relies on external wafers, read manufactured by TSM, not Intel. Intel also needs to buy memory. It's including on each chip as Lunar Lake laptops don't have separate memory sticks. Those are reasons why Lunar Lake will only modestly improve the company's situation in 2025, he says. Intel says it's now restructuring, suspending its dividend, and spending less period, but will, quote, maintain its core investments to execute its strategy and build a resilient and sustainable semiconductor supply chain in the U.S. and around the world, end quote. So this all sounds like their big foundry initiative to reinvent themselves,

is costing them a lot of money, so much so that they're going to have to cut to the bone. It sounds like they're basically burning anything flammable just to keep warm while they wait for a turnaround. And while everybody else is spending all they can to build out and meet the AI moment, they're having to come out and say they're reducing spending. That is not what the street wants to hear right now. Speaking of Companies in Trouble, a special committee of 23 and Me's board says it's not, quote, prepared to move forward with CEO Anne Wajicki's 40 cents per share offer to take the company private, quoting Bloomberg. On Monday, Wajiki proposed buying the outstanding shares of 23 and Me for 40 cents each, according to a filing this week. In a letter to Wajiki, the independent committee formed to explore deals, said that it was, quote, disappointed with the proposal because it wasn't at a premium to Wednesday's close, which was 40 cents a share and lacked financing commitments.

Wuziki, whose chair of the board indicated in April that she planned to maintain control of the company and would, quote, not be willing to support any alternative transaction. The special committee is asking her to withdraw her opposition to outside buyers and said in a letter made public Friday that it's moving forward with a, quote, revised business plan or other alternatives, regardless of whether Wizhiki offers a higher offer price. Given both the lack of certainty regarding a path forward with you and your potential investors and the current liquidity position of the company, the board wrote that even while she's, she's considering making a better offer, quote, we intend to immediately begin the process of engaging a consultant to advise the special committee on a revised business plan that would provide the company with a path to a more sustainable financial profile and achieving profitability, end quote.

When the company agreed to go public in 2021, it was valued at $3.5 billion. Since then, sales of DNA testing kits have slowed, and the company has pivoted to offering subscription products in hopes of creating repeat customers for its consumer business. So far, that business hasn't generated the number of signups the company had initially anticipated. The stock has traded below a dollar a share since late last year, and it has until November to raise share prices to stay compliant and remain listed. The stock is down nearly 60% this year, end quote, and just checking the market cap this morning, $193 million. Yet another new AI startup and another new AI model, but this one is interesting to me for a couple of reasons.

Black Forest Labs is a new startup from the original founders of stable diffusion, so that's interesting point number one. They've secured $31 million for a new AI image generator called Flux 1. And the second thing that's interesting to me is the novelty of this model. People have been asking what comes next beyond transformer models and attention. Well, maybe we're starting to see some hints. Listen to these tweaks, quoting Venture Beat. Flux 1 introduces several technical innovations. The models employ flow matching, a method that generalizes diffusion models

and incorporate rotary positional embeddings and parallel attention layers, for enhanced performance and hardware efficiency. This approach has yielded impressive results in visual quality, prompt adherence, and output diversity. The impact of Flux 1 could extend far beyond the AI research community. Graphics designers, digital artists, and creative professionals may discover new possibilities in the model's ability to generate high-quality images across a wide range of styles and aspect ratios. Flux 1 comes in three variants, the closed-source flux 1 Pro available via API, the open-weight flux 1 dev for non-commercial use,

and Flux 1 Schnell, a faster version released under an Apache 2.0 license for personal and local development. All models boast an impressive 12 billion parameters and utilize a hybrid architecture of multimodal and parallel diffusion transformer blocks. Early demonstrations suggest flux 1's output quality rivals or even surpasses that of popular closed source models like Mid-Journey 6 and Dolly 3. The timing of Flux 1's release proves crucial for open source AI, recent turbulence at stability AI, company behind Stable Defusion had raised concerns about the future of accessible high-quality image generation models. Black Forest Labs' entry into the space could reinvigorate the open-source AI ecosystem and potentially accelerate innovation across various fields, from graphic design to scientific visualization. Additionally, the open nature of Flux 1 dev and Schnell

variants could spark a new wave of applications and integrations across various industries. Black Forest Labs has already set its sights on its next frontier, state-of-the-art text-video systems. Success in this arena could further cement the company's position as a leader in generative media technology, end quote. Time for the weekend long-rate suggestions. First up from Wired, a look at Ray V. Light, an open-source laser hacking tool built by two hackers for less than $500 that can reverse engineer chips. Quote, at the Blackap Cybersecurity Conference in Las Vegas next week, Sam Beaumont and Larry Patch Trowell, both hackers at the security firm NetSPI, planned to present a new laser hacking device they're calling the Ray V Light. Their tool, whose design and component

lists they plan to release open source, aims to let anyone achieve arcane laser-based tricks to reverse-engineer chips, trigger their vulnerabilities and expose their secrets, methods that have historically only been available to researchers inside of well-funded companies, academic labs, and government agencies. State-of-the-art commercial tools for light-based hacking techniques, such as risk-ure laser station have typically costs as much as $150,000, and even lower-budget versions cost closer to $10,000, yet through a combination of 3D printing, commodity component choices and clever physics tricks, Beaumont and Trowell built theirs for less than 500 bucks. Their goal in creating and releasing the designs for that ultra-cheap chip hacking gadget,

they say, is to make clear that laser-based exploitation techniques known as laser fault injection or laser logic state imaging are far more possible than many hardware designers, including clients for whom Beaumont and Trowell sometimes perform security testing at NetSPI believe them to be. By demonstrating how inexpensively, those methods can now be pulled off, they hope to both put a new tool in the hands of DIY hackers and researchers worldwide, and to push hardware manufacturers to secure their products against an obscure but surprisingly practical form of hacking. Beaumont describes the Ray V. Light as part of a larger trend she calls the domestication of tooling. Devices like the Chip Whisperer and HackRF have made

electromagnetic or radio-based hacking techniques vastly cheaper and more accessible. The RayV. Light, she hopes will do the same for lasers. It's significant, says Adam Lari, a long-time hardware hacker and current head of product security at Electric Vehicle Charging Firm Alpitronic, who reviewed Beaumont and Trial's Laser hacking work. It moves the tools from the super-expensive academic or state actor platform to the garage where the really inventive stuff happens, he said. And finally from The Verge, an Olympic theme story about the AI-assisted judging support system for scoring in gymnastics. Quote, In its early days, the system used LiDAR, light detection and ranging technology to create 3D composites of gymnasts in action.

These days, it uses an even more sophisticated system drawing from 4 to 8 strategically placed high-deaf cameras to capture the movement of the athletes making 3D models and identifying whether the elements they are performing fall into the parameters established by the judging bodies inside the Federation. But the computer system doesn't make judgments itself. Instead, it is deployed when there is an inquiry from the gymnast or coaches or, on a dispute within the judging panel system itself. The judging support system or JSS can be consulted to calculate the difficulty score of an athlete's exercise, a second opinion rather than an initial prognosis. Currently, it is mostly used for edge cases. The JSS wasn't necessary to evaluate

the value of Biles's vault at the World Championships in Antwerp. Her performance on that vault was too emphatic to be borderline. Still, the cameras positioned at the corners of the vault podium captured her 3D likeness as they did for all of the other athletes. who competed through the 2023 World Gymnastics Championships. The technology distilled the legendary athlete and her performance down to straight lines and sharp edges. It showed the distance and height she traveled in numbers. The awe and wonder one feels when watching Biles perform could now be recognized by a computer, understood, though not exactly appreciated.

Fujitsu and Figg announced JSS back in 2017 with the goal of having a system up and running by the Summer Olympics in 2021. A home games in Tokyo would have been an ideal opportunity for the Japanese-based tech conglomerate to showcase this kind of technology, and it would have been a noteworthy achievement for Moriani Wanatabi, the first Japanese president of the Luzon-based fig. But the JSS wasn't ready. In fact, it would take another four years of work. At the 2023 World Championships in Antwerp, the JSS was finally ready to go on all 10 artistic gymnastics apparatuses, six for the men and four for the women, end quote.

Nothing for you this weekend in terms of bonus episodes. Talk to you again on Monday.