AI Infrastructure Boom: Meta's Piracy Controversy and CoreWeave's IPO Reveal Market Risks

[00:00:02] It's Monday, March 24th, 2025 and I'm Dave Solt. Three things to know today. Clouds rise, max fly, and the job market sighs. A look at the state of play. Meta's alleged book piracy raises IP risks across the generative AI ecosystem and CoreWeaves IPO balances sky-high growth and soaring costs as Microsoft rethinks their AI spend. This is the Business of Tech.

[00:00:28] Global investments in data centers surged by 51% last year, reaching $455 billion, according to a report by the Delorio Group. The firm anticipates a further 30% growth in infrastructure capital expenditures this year. The majority of the spending was directed towards expanding cloud vendor capacities, with the top 10 hyperscalers accounting for over half of the global capital expenditures in 2024.

[00:00:55] Notably, companies like Amazon Web Services, Microsoft and Google Cloud are heavily investing in infrastructure to meet the rising demand for artificial intelligence capabilities. Delorio predicts that annual capital expenditures could exceed $1 trillion by 2029. Additionally, enterprises increase their hardware and infrastructure spending by 25% year-over-year, totaling $150 billion in 2024, marking the largest growth rate since the

[00:01:25] firm began tracking the market in 2014. The U.S. The U.S. PC market is projected to see a modest growth of 2% this year, amid ongoing tariff concerns affecting prices for consumers and businesses. According to data from Canalys, PC shipments to the U.S. Hello? pitcher, McDonald's penny, oil-over-year-old-of-year, FX-plus. The U.S. Office of the U.S. All right.

[00:01:53] In what you see here, the U.S. Obviously, the U.S. The U.S. world-day most recently economy-over-year economy-over-of-year, sodas have so many apps, but The U.S. A. All right. On the right. The U.S. A. Tr GREGOR as of January 2025, compared to 29% in November 2024. The commercial segment has been particularly

[00:02:19] strong, spurred by the approaching end-of-service date for Windows 10, prompting businesses to refresh their PC fleets. However, the tariffs had led to uncertainty in pricing and have spurred stockpiling within the industry. Notably, Apple achieved a milestone by surpassing Lenovo in the vendor rankings, with a 26% annual growth in max shipments during the same quarter. Axios looked

[00:02:46] at the trend of job growth in the U.S., which has become increasingly reliant on government, healthcare, and education sectors. According to Treasury Secretary Scott Bessent and other officials from the Trump administration, this dependency indicates a weaker economy than it appears. Over the past two years, employment in government and healthcare has increased by 3.2 million jobs, a 6.7% rise, while other sectors collectively added only 948,000 jobs, a mere 0.9% increase.

[00:03:15] This shift raises concerns about future labor market stability, as hiring in these sectors is expected to slow significantly, from 100,000 jobs per month now to just 15,000 in the latter half of the year, according to Goldman Sachs economists. The thesis? That the labor market's reliance on low productivity sectors may hinder economic growth and challenges the sustainability of this current

[00:03:39] trajectory. Why do we care? Most of the data center CapEx is consolidated among the top 10 hyperscalers, making it a high-stakes, capital-intensive game that many players simply can't afford. The enterprise segment also saw a significant 25% year-over-year increase in hardware and infrastructure spending, indicating that AI workloads and data locality concerns are pushing even traditional enterprises to invest in on-prem or hybrid environments. That projected 30% year-over-year

[00:04:09] CapEx growth with long-term forecasts of $1 trillion carries bubble-like signals, especially if AI monetization fails to keep pace. Infrastructure buildouts don't guarantee ROI, particularly if generative AI solutions under deliver commercially or face regulatory friction. With over 50% of global CapEx from just 10 companies, any pushback by the hyperscalers could sharply deflate market momentum.

[00:04:36] This isn't a broad-based boom. It's highly top-heavy. PC shipments in the U.S. rose 6% year-over-year in Q4 of last year, with projected 2% growth this. The catalyst? Windows 10 end-of-life driving commercial refresh cycles. However, 50% of channel partners sitting on 5-plus weeks of inventory, up from 29% in November 2024. It's not sustained demand. It's a deadline-driven refresh cycle.

[00:05:03] The spike is likely front-loaded in this 2024-2025 time period, creating a potential hangover effect next year. The glut of inventory also suggests that vendors and resellers overestimated demand, likely due to tariff fears and supply chain uncertainty. Private sector job growth is stagnant, meaning tech service demand from non-public sectors may underwhelm unless enterprise and

[00:05:28] S&B investments rebound. If job growth continues to skew towards sectors with traditionally lower productivity, like government, education, and healthcare, this could suppress GDP growth and technology investment. Especially for firms offering productivity enhancement solutions, ROA cases may be harder to make. Customers operate in budget-constrained, regulation-heavy sectors.

[00:05:53] With every new breach and threat that I cover, it's clear that cybersecurity isn't a luxury anymore. It's a necessity. That's where Huntress comes in. Their fully managed cybersecurity platform is built for every kind of business, not just the 1%. Huntress seamlessly integrates their products and threat hunting team. Their EDR, ITDR, SIM, and security awareness training solutions are purposely built for their elite 24x7

[00:06:20] security operations center to stop threats before anyone else even spots them. This potent combination of purpose-built cybersecurity and threat hunting expertise is one of the many reasons why G2 users have voted Huntress the number one rated EDR for growing businesses. To see what people-powered cybersecurity looks like,

[00:06:40] visit Huntress.com slash MSB Radio. The story of the end of last week was about how Meta pirated books to create its AI. A recent investigation by The Atlantic highlights the extensive use of pirated books by Meta to train its AI model, Lama 3. Court documents reveal that Meta employees expressed urgency in acquiring books,

[00:07:05] stating that they are, quote, actually more important than web data, end quote. To expedite this process, Meta turned to Library Genesis, a massive repository of over 7.5 million books and 81 million research papers, which is known for its illegal distribution of copyrighted material. Despite legal risks, Meta reportedly engaged in torrenting to access this vast library. The implications of using such pirated

[00:07:32] resources raise significant ethical questions about the future of knowledge sharing, as generative AI technologies increasingly rely on copyrighted works without proper licensing. Why do we care? This is not just about one company's ethical lapse. This is about the foundational legality of large-scale generative AI development. The strategic implication is massive. Training data provenance is now a commercial

[00:07:57] liability. Enterprises will not be able to blindly trust foundation models for commercial or regulated use cases if there's no clear IP chain of custody. A two-tier AI ecosystem will form, one that plays by copywriting and licensing rules and moves slower, and one that scrapes, pirates, and ships fast. Service providers and consultants need to evaluate the AI models they integrate or resell. Clients in regulated

[00:08:25] sectors like finance, healthcare education, and more could face legal exposure by proxy if the underlying model includes illicit training data. Vetting model providence, once a technical curiosity, is a due diligence requirement. Meta positioned Lama as the open alternative to OpenAI and Google. But here's the problem. If open-weight models are built on illegally sourced data, they are no longer safely reusable.

[00:08:52] Companies that value open models for flexibility or cost have to factor in IP exposure as a hidden cost. I also wanted to talk about CoreWeave. So first, the basics. CoreWeave, a cloud AI startup backed by NVIDIA, is planning to raise up to $2.7 billion through its initial public offering, which could give the company a market valuation of $26 billion. The company is offering shares priced between $47 and $55

[00:09:21] each, with expected net proceeds of approximately $2.3 billion. OpenAI will receive $350 million worth of CoreWeave stock as part of the deal. Notably, Microsoft accounts for 62% of CoreWeave sales, which generated about $1.9 billion in revenue in 2024, marking a 737% increase from the previous year.

[00:09:45] CoreWeave plans to trade under the symbol CRWV on the NASDAQ, and the company is expanding its presence in Europe, committing $2.2 billion to open new data centers across Sweden, Norway, and Spain by the end of the year. And now, the but. Microsoft has chosen not to exercise a nearly $12 billion option to purchase additional data center capacity from CoreWeave, signaling a shift in how companies are

[00:10:13] managing their artificial intelligence budgets. CoreWeave has quickly secured another buyer, with OpenAI acquiring the contract last week. Despite the decision, Microsoft reiterated its commitment to invest $80 billion in AI. The company is being more strategic about its spending, according to CEO Satya Nadella, who stated on CNBC that they remain committed to their budget. The information is reporting that CoreWeave is facing a reality check as it prepares for its IPO.

[00:10:40] Initially, CoreWeave projected revenue growth to quadruple to $8 billion in 2025, with a cash burn reduction to $4 billion. However, new forecasts from analysts at major investment banks, including Goldman Sachs and JPMorgan Chase, indicate that revenue may only reach approximately $4.6 billion this year, with cash burn rising significantly from $6 billion last year to about $15 billion this year.

[00:11:07] These adjustments reflect a 40% reduction in growth expectations compared to last fall. CoreWeave is starting its investor roadshow, aiming to raise a smaller amount of money than anticipated amidst a volatile stock market, but still expects a valuation around $30 billion in the midpoint of its range. The company is under pressure to meet these more conservative projections, especially given it's one of the fastest growth rates for companies going public in recent years.

[00:11:33] CoreWeave's IPO could be one of the largest tech listings in recent history, with plans to sell about $2.5 billion in stock. Why do we care? CoreWeave is a pure-play proxy for the AI infrastructure boom, and its IPO will be a litmus test for the market's appetite for AI scale-up bets. But the fundamentals paint a familiar picture. A company scaling revenue off a single hyperscaler customer, Microsoft, operating with outsized cash burn that

[00:12:01] dwarfs current revenues, in an industry that's rapidly learning that not all AI bets are profitable or even sustainable. For IT service providers and cloud integrators, don't assume AI infrastructure demand is infinite or evenly distributed. CoreWeave's story is a warning about customer concentration risk, shifting capex priorities, and volatility of AI fuel growth narratives. The IPO will be one of the biggest

[00:12:27] tech listings and its reception will signal whether public markets still buy into the picks-and-shovels narrative of AI infrastructure. If it succeeds, it could unlock a wave of follow-on IPOs from data center and cloud infrastructure players. It may validate the case for investing in AI-enabled infrastructure, even if application layer ROI is still murky. And if it stumbles, we'll see a market

[00:12:52] correction in AI cloud valuations, especially those with unproven revenue durability, and it could trigger a second-order rethink of how much AI infrastructure is really needed, and how much demand is durable versus speculative. Are you and your clients tired of the time-consuming ticket tennis of coordinating meetings and help desk calls? If you're looking at the time-consuming ticket, you're looking at the time-consuming ticket to the time-consuming ticket

[00:13:20] that connects directly to ConnectWise Manage or Autotask? TimeZest offers scheduling automation that gives you complete control of your schedule and eliminates the hassle of calendar ping-pong. As the only service designed specifically for MSPs, it integrates into your workflow and makes scheduling appointments easy on you and your clients. Plus, you can try TimeZest for free. Visit

[00:13:48] TimeZest.com slash MSBRadio and use the code MSBRadio to get 10% off your first year of TimeZest. Thanks for listening. Today is National Cheese Steak Day, National Chocolate-Covered Raisin Day, National Cocktail Day, and International Day for Achievers. You go. Nerdy Ocon will be held in Palm Springs, California from April 7th through 9th. Visit

[00:14:16] NerdyOcon.com to learn all about it. The Business of Tech is written and produced by me, Dave Sobel, under ethics guidelines posted at businessof.tech. If you've enjoyed the show, make sure you've subscribed or followed on your favorite platform. It's free and helps directly. Give us a review, too. If you want to support the show,

[00:14:38] visit patreon.com slash MSP Radio and you'll get access to content early. Or buy our Why Do We Care merch at businessof.tech. Have a question you want answered? We take listener questions, send them in, ideally as a voice memo or video to question at MSPradio.com. I answer listener questions live on our Wednesday live show on YouTube and LinkedIn. If you've got a comment or a thought on a story, put it in the comments if you're on YouTube or reach out on

[00:15:08] LinkedIn if you're listening to the podcast. And if you want to advertise on the show, visit MSPradio.com slash engage. Once again, thanks for listening and I will talk to you again on our next episode. Part of the MSP Radio Network.