Tech Brew Ride Home - Mon. 11/25 – Return Of The Vita?
Episode Date: November 25, 2024Is Sony about to get back into the mobile gaming hardware business? Is Bluesky in the dog house with the EU? Nvidia’s new AI model. Is Intel not going to get as much money as it hoped for? And the r...ise of AI superclusters. Sponsors: MackWeldon.com promocode BRIAN 1Password.com/ride Links: Sony Working on Handheld Console for PS5 Games to Rival Switch (Bloomberg) Bluesky breaching rules around disclosure of information, says EU (Financial Times) Nvidia shows AI model that can modify voices, generate novel sounds (Reuters) NVIDIA's new AI model Fugatto can create audio from text prompts (Engadget) Washington Curtails Intel’s Chip Grant After Company Stumbles (NYTimes) Klarna’s Planned IPO Sets the Stage for More Fintech Listings (Bloomberg) Exclusive: CoreWeave targets valuation of over $35 billion in 2025 US IPO, sources say (Reuters) AI’s Future and Nvidia’s Fortunes Ride on the Race to Pack More Chips Into One Place (WSJ) Learn more about your ad choices. Visit megaphone.fm/adchoices
Transcript
Discussion (0)
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 Monday, November 25th, 2024. I'm Brian McCullough. Today is Sony about to get back into the mobile gaming hardware business?
Is Blue Sky in the doghouse with the EU? Invidia's new AI model? Is Intel not going to get as much money as it hoped for? And the rise of AI superclusters. Here's what you miss today in the world of tech.
Hey, it looks like at long last we might finally get a successor to the good old Vita. Sources say Sony is in the early
stages of developing a portable console to play PS5 games, building on the PlayStation Portal,
although its launch is still probably years away. Quoting Bloomberg,
the idea builds on the PlayStation Portal, an 8-inch handheld device Sony released in
2023 that allows users to play PS5 games by streaming them over the internet.
Tokyo-based Sony initially had intended for it to function as a standalone device like
Valve's Steam Deck, the people said. Building a mobile device that's capable of directly
playing PS5 games would potentially make Sony's software more accessible and appealing to a wider
audience. The company has taken several steps in recent times in that direction, including making
a bigger push into mobile and PC gaming as well as live service titles. Mobile gaming,
which brings in the bulk of the industry's revenue each year, is today dominated by smartphones.
Nintendo's Switch console has carved out a lucrative niche for itself by offering the
versatility of being playable both on the move and connected to a TV set. Its successor is expected to be
released in 2025 and will retain compatibility with Nintendo's rich library of existing Switch
titles, potentially sparking an upgrade cycle among users of the nearly eight-year-old console.
Sony and Microsoft have not competed with their own portable gaming hardware for years,
following popular but ultimately unsuccessful products like the PSVita and PlayStation Portable,
which were discontinued a decade ago, end quote.
This is sort of out of left field.
The EU says Blue Sky is in breach of its rules for not.
not disclosing key details about itself. They've asked 27 governments to see, quote, if they can find
any trace of blue sky. Quoting the F.T, all platforms in the EU have to have a dedicated page on their
website where it says how many users they have in the EU and where they are legally established,
said Commission spokesperson Thomas Reganer. This is not the case for blue sky as of today.
This is not followed, end quote. The intervention comes as thousands of users, including
Commission President Ursula von der Leyen have opened Blue Sky accounts in recent weeks.
Renniger said the commission the EU's executive arm had written to the 27 national governments
to see, quote, if they can find any trace of blue sky, such as identifying an EU-based office.
It has not yet contacted the company directly, he added.
Blue Sky is a U.S. public benefit company led by digital rights activist and software engineer Jay Graber.
It was founded in 2019 for the purpose of developing a single standard or protocol upon which
social platforms and other developers could build their own operations. It has come to resemble
X with users posting short messages and images. The EU move comes as regulators are increasing
scrutiny over potential breaches of the digital rules by Musk's X. Brussels has been investigating
his social media platform since last year over accusations of noncompliance with its landmark
Digital Services Act, mainly in relation to the alleged spread of illegal content and
misinformation. Fresh probes on other platforms are also likely. Last month, EU regulators sent questions
and requests for official documents to YouTube, TikTok, and Snap on how the online platforms are monitoring the use of artificial intelligence in their algorithms over concerns that it could lead to illegal and fake content online.
Companies found in breach of the law face penalties of up to 6% of their global annual revenues, repeat offenders could even be banned from the block.
The Commission cannot regulate blue sky directly as it does not yet reach the threshold of more than 45 million monthly users in the EU to be designated a very large online platform.
But Reniger says that if member states could identify an EU-based representative for the company,
Brussels would, quote, reach out to blue sky, end quote.
NVIDIA has unveiled Fugato, a new AI model for generating music and audio that can modify voices and make novel sounds,
trained on open source data, quoting Reuters.
Invidia, the world's largest supplier of chips and software used to create AI systems, said it does not have immediate plans to publicly release the technology,
which it calls Fugato short for foundational generative audio transformer Opus 1.
The new model generates sound effects and music from a text description, including novel sounds
such as making a trumpet bark like a dog.
What makes it different from other AI technologies is its ability to take in and modify existing audio,
for example by taking a line played on a piano and transforming it into a line sung by a human voice,
or by taking a spoken word recording and changing the accent used and the mood expressed.
If we think about synthetic audio over the past 50 years, music sounds different now because of computers, because of synthesizers, said Brian Katanzaro,
vice president of applied to deep learning research at Nvidia. I think that generative AI is going to bring new capabilities to music, to video games, and to ordinary folks that want to create things, end quote.
And quoting and gadget.
Invidia calls it a Swiss Army knife for sound. The company listed some possible real-world scenarios wherein Fugato could be of use in its announcement.
Music producers, it suggested, could use the technology to quickly generate a prototype for a song idea,
which they can then easily edit to try out different styles, voices, and instruments.
People could use it to generate materials for language learning tools in the voice of their choice,
and video game developers could use it to create variations of pre-recorded assets to fit changes in the game
based on the player's choices and actions.
In addition, the researchers found that the model can accomplish tasks not part of its pre-training with some fine-tuning.
It could combine instructions that it was trained on separately, such as generating speech that sounds angry with a specific accent or the sound of birds singing during a thunderstorm.
The model can generate sounds that change over time as well, like the pounding of a rainstorm as it moves across the land, end quote.
Services say Intel, the biggest recipient of Chips Act money will see its funding drop to less than $8 billion from the $8.5 billion that the U.S. announced earlier in 2024.
It doesn't sound like much on first blush, but that is $500 million, and given the situation
Intel is in lately, they probably need every dime, you would think, quoting the Times.
The change in terms takes into account a $3 billion contract that Intel has been offered to produce
chips for the U.S. military, two of these people said.
The government's decision to reduce the size of the grant follows Intel's move to delay
some of its planned investments in chip facilities in Ohio.
The company now plans to finish that project by the end of the decade instead of 2025.
The chipmaker has been under pressure to reduce costs after posting its biggest quarterly loss in the company's 56-year history.
The move by the Biden administration also takes into account Intel's technology roadmap and customer demand.
Intel has been working to improve its technological capacity to catch up to rivals like Taiwan Semiconductor,
but it has struggled to convince customers that it can match TSM's technology.
Intel's troubles have been a blow to the Biden administration's plans to rev up the domestic chip manufacturing.
In March, President Biden traveled to Arizona to announce Intel's multi-billion dollar award
and said the company's manufacturing investments would transform the semiconductor industry.
Intel's investment was at the forefront of the administration's ambition to return chip manufacturing
to the United States from Asia.
The Chips Act, a bipartisan bill passed in 2022, provided $39 billion in funding to subsidize
the construction of facilities to help the United States reduce its reliance on foreign
production of the tiny critical electronics that power everything from iPads to dishwashers.
To prevent taxpayer money from being wasted, Commerce Department officials set milestones for the
companies to meet in order to get funds. The benchmarks included building a plant, producing chips,
and signing up customers to buy domestically produced products. Intel, which lobbied aggressively for
the bill's passage, was long seen as the law's biggest beneficiary, but its business struggles
complicated its negotiations over its final award. The Commerce Department had granted the initial
award to support Intel's expansion of operations in Arizona, New Mexico, and Oregon,
as well as the construction of two plants in Ohio, end quote.
Something that put on our collective radars for next year, IPOs. Yes, IPOs might be back and soon.
In the run-up to an expected IPO of its own, Clarna reported around $1.85 billion in Q1 to Q3-2020 revenue, up 23% year-on-year.
A pre-tax lost down 99% to only around $180,000, and Q3 net income up 57% year-percent,
year-on-year to around $19.7 million. If Klarna does go out the door sometime in the spring,
it could lead the way for other fintechs like Trusley, Chime, Zilch, Plaid, Revolut, Zopa,
Brex, and Ramp. Quoting, quoting Bloomberg, how fast these companies come to market and where
they choose to list could be influenced by how Klarna fares. Clarner's much-anticipated filing could
presage an uptick in fintech public offerings after a relatively slow period, said Mark Palmer,
senior research analysts for fintech and digital assets at the benchmark company.
The recent upward surge in fintech stock prices and valuations bodes well for new public
offerings in the space, as do investors' expectations of a more fintech-friendly regulatory
regime during the new Trump administration, end quote.
Klarna, Zilch, Trusley, Trusely majority owner, Nordic Capital, Revolut, Plad,
Brex, and Shime declined to comment or did not respond to requests for comment for the story.
A Zopa Banks boastsperson said, quote, an IPO is not.
an immediate priority, but that it continues to work toward one, preferably in the U.K., when the
time is right. A ramp spokesperson said, we have ambitions to be a public company, but aren't
actively planning for it at the moment, end quote. Klarna confidentially filed for an IPO with the U.S.
Securities and Exchange Commission, it said in a statement last week, while the company provided
no financial details analysts last month put Klarna's implied valuation at about $14.6 billion
after shareholder Kristallis investments increased the value of its stake. That would mark
an improvement from the $6.7 billion valuation it garnered in its last funding round in 2022,
but is still much lower than the $45.6 billion valuation Klarna boasted in 2021.
The industry needs liquidity and valuations need to come back to Earth, said Sidney Thomas,
a founding general partner at Symphonic Capital.
She called Clarnas private valuation decline reasonable and its current estimated valuation,
quote, still exceptional, given what the industry has been through.
I hope Klarno's IPO gives other founders the courage to pursue an exit, even if it means doing so at a lower valuation, Thomas said, end quote.
But there's another one to be on the lookout for two, Coor-Weave, which could be the first big IPO of this AI era, and big just generally, as they are apparently aiming for a valuation of over $35 billion.
They're likely to target raising more than $3 billion if it does go public around the second quarter of next year, quoting Reuters.
Funding for private AI and cloud startups in the U.S., Europe, and Israel is rising after three years of decline
and is estimated to touch $79.2 billion by the end of this year, venture capital firm Assal said in October.
Corweave offers access to data centers and high-powered chips for AI workloads,
primarily supplied by Nvidia, one of the company's main backers.
It competes against larger cloud computing service providers such as Microsoft and Amazon's AWS.
U.S. Corweave recently completed a $650 million secondary share sale, which valued it at $23 billion, investors led by Jane Street, Magnitar, Fidelity Management, and McQuary Capital participated in the secondary stock deal. Bloomberg reported in November that Corweave has tapped investment banks for its IPO preparations, end quote.
Speaking of, the journal had an interesting piece this weekend, taking a look at the rise of so-called superclusters that use around 100,000 Nvidia GPUs for,
training giant AI models and the new engineering challenges arising from such clusters.
Quote, Elon Musk's XAI built a supercomputer it calls Colossus with 100,000 of
Nvidia's Hopper AI chips in Memphis in a matter of months.
Meta chief executive Mark Zuckerberg said last month that his company was already training
its most advanced AI models with a conglomeration of chips he called,
bigger than anything I've seen reported for what others are doing.
A year ago, clusters of tens of thousands of...
of chips were seen as very large. Open AI used around 10,000 of Nvidia's chips to train the
version of ChatGPT. It launched in late 2022. UBS analysts estimate a push toward larger superclusters
could help Nvidia sustain a growth trajectory that has seen it rise from about $7 billion
of quarterly revenue two years ago to more than $35 billion today. That jump has helped make it
the world's most valued publicly listed company with a market capitalization of more than
three and a half trillion. The continuation of the AI boom for Nvidia also
depends in great measure on how the largest clusters of chips pan out. The trend promises not only
a wave of buying for its chips, but also fosters demand for Nvidia's networking equipment,
which is fast becoming a significant business and brings in billions of dollars of sales each year.
Nvidia chief executive Jensen Huang said in a call with analysts following its earnings
Wednesday that there was still plenty of room for so-called AI Foundation models to improve
with larger-scale computing setups. Huang said that while the biggest clusters for training
for giant AI models now top out at around 100,000 of Nvidia's current ships, quote,
the next generation starts at around 100,000 Blackwells. And so that gives you a sense of where
the industry is moving. Wang marveled on a podcast last month at the speed with which Musk had built
his colossus cluster and affirmed that more larger ones were on the way. He pointed to efforts to
train models distributed across multiple data centers. Do we think that we need millions of GPUs?
No doubt, Wong said. That is a certainty now. And the question is, how do we
architect it from a data center perspective. Unprecedented superclusters are already getting airplay.
Musk posted last month on his social media platform X that his 100,000 chip colossus supercluster
was soon to become a 200,000 chip cluster in a single building. He also posted in June that
the next step would probably be a 300,000 chip cluster of Nvidia's newest chips next summer.
The rise of superclusters comes as their operators prepare for the Blackwell chips,
which are set to start shipping out in the next couple of months. They are estimated to cost around
$30,000 each, meaning a cluster of $100,000 would cost $3 billion, not counting the price of the
power generation infrastructure and IT equipment around the chips.
Those dollar figures make building up superclusters with ever more chips, something of a gamble
industry insiders say, given that it isn't clear that they will improve AI models to a degree
that justifies their cost.
New engineering challenges also often arise with larger clusters.
Meta researchers said in a July paper that a cluster of more than 16,000 of Nvidia's GPUs
suffered from unexpected failures of chips and other components routinely as the company trained
in advanced version of its Lama model over 54 days.
Keeping Nvidia's chips cool is a major challenge as clusters of power-hungry chips become packed
more closely together, industry executives say, part of the reason there is a shift
toward liquid cooling where refrigerant is piped directly to chips to keep them from
overheating.
And the sheer size of the superclusters require a stepped-up level of management of those
chips when they fail. Mark Adams, chief executive of Penguin Solutions, a company that helps set up
and operate computing infrastructure, said elevated complexity in running large clusters of chips inevitably
throws up problems. When you look at everything that can go wrong, you could be utilizing
half of what your capital expenditure was because of all of these things that can break down,
he said, end quote. But the market still seems to think this is where the action is. I was reading over the
weekend about the biggest stock on the Niki index this year. Apparently, it's Japan's Fujifax.
which is up more than 400% in 2024. What does it do? It specializes in fiber optic cable for,
you guessed it, AI data centers. Nothing more for you today. Talk to you tomorrow.
