💸 CoreWeave secures $8.5 billion loan to expand AI infrastructure ↗
CoreWeave landed a huge new financing package to keep feeding the compute beast. The company said the delayed-draw facility will support the expansion of its AI cloud platform, which feels perfectly in step with this phase of the market - more GPUs, more data centers, more debt, more of everything.
It can initially borrow about $7.5 billion, with a step-up to $8.5 billion as assets stabilize. Morgan Stanley and MUFG co-structured the deal, Goldman Sachs and JPMorgan joined in, and Blackstone Credit & Insurance anchored it - a clear sign this is no longer a scrappy startup tale. (Reuters)
🧠 Nvidia bets $2 billion on Marvell as rising AI adoption fuels competition ↗
Nvidia put $2 billion into Marvell, and the subtext is fairly plain - stay central, even as more customers drift toward custom chips. It reads as defensive and expansive at the same time.
The deal ties Marvell’s custom silicon and networking gear into Nvidia’s ecosystem, with a focus on optical interconnects and silicon photonics. Put simply: faster, more power-efficient data movement inside giant AI systems, which has become one of the genuine bottlenecks, not just a contest over raw chip bragging rights. (Reuters)
💬 Salesforce announces an AI-heavy makeover for Slack, with 30 new features ↗
Slack got a big AI gloss-over - or a glow-up, depending on your tolerance for enterprise software demos. Salesforce unveiled 30 new features, with the biggest push centered on making Slackbot more agent-like and more capable across everyday work.
The standout addition is reusable AI skills, which let people define repeatable tasks that Slackbot can run across channels, apps, and company data. Salesforce is pitching a future where you type something simple like “create a budget,” and the bot handles the scavenger hunt, the draft, and even the meeting setup. A lot, perhaps. Still, that is kind of the point. (TechCrunch)
🛡️ AI companies that want to work with the state of California will have to meet new privacy and security standards. ↗
California moved to tighten the screws on AI vendors that want state business. That matters because, while Washington keeps wobbling around broader AI rules, states are still finding ways to set practical guardrails where procurement is concerned.
The Verge reports the new standards come via an executive order from Governor Gavin Newsom. So this is less about flashy frontier-model rhetoric and more about a familiar lever - if you want government contracts, meet the privacy and security bar first. It is not glamorous, but it tends to stick. (The Verge)
🚪 Yupp shuts down after raising $33M from a16z crypto’s Chris Dixon ↗
Yupp, the startup that let users compare outputs from hundreds of AI models and give feedback on what worked best, is shutting down less than a year after launch. That is a rough comedown for a company with a big seed round and several very shiny names around it.
The founders said the product never found strong enough fit, even though Yupp had signed up 1.3 million users and was collecting millions of preference signals a month. The twist is almost too on-the-nose: AI moved so fast that the thing built to help evaluate models got outrun by the models themselves. (TechCrunch)
🍎 Apple tests Siri feature that handles multiple commands at once, Bloomberg News reports ↗
Apple is reportedly testing a Siri feature that can process multiple requests in a single query, which sounds basic for modern AI assistants, and yet for Siri it marks a telling shift. Apple seems to be inching from assistant toward chatbot, though it may not call it that just yet.
Reuters says the feature is being built into upcoming Apple operating systems, while the broader Siri revamp is tied to Gemini tech and expected to be shown at WWDC on June 8. The larger read is simple - Apple still looks like it is catching up, but it is finally moving with greater urgency. (Reuters)
⚡ Big Tech's $635 billion AI spending faces energy shock test, S&P Global says ↗
One of the bigger AI stories was not a product launch at all - it was the bill. S&P Global warned that the giant AI infrastructure spree planned by Microsoft, Amazon, Alphabet, and Meta could be stress-tested by rising energy costs and broader geopolitical instability.
The key number is wild enough on its own: $635 billion in planned AI-related spending this year, up from $383 billion the year before. The catch is that data centers are power-hungry little furnaces, so if energy stays expensive, the AI boom starts to look less like a rocket and more like a shopping trolley with one broken wheel. (Reuters)
FAQ
Why is AI infrastructure getting so much new financing right now?
AI infrastructure is drawing heavy financing because demand for compute, GPUs, and data center capacity keeps climbing. In this roundup, CoreWeave’s multibillion-dollar loan signals that investors still believe large-scale AI cloud demand has room to grow. The broader pattern is straightforward: companies want more capacity now, even when that means taking on substantial debt to build more quickly.
What does CoreWeave’s $8.5 billion loan mean for the AI market?
It suggests AI infrastructure is no longer funded like a speculative side bet. A facility of that size, backed by major banks and credit firms, points to a more mature and capital-intensive phase of the market. It also shows that expanding AI cloud platforms now depends as much on financing structure and asset stability as on technical ambition.
Why did Nvidia invest in Marvell instead of just building everything itself?
The move appears to be a way for Nvidia to remain central as more customers explore custom chips and specialized networking. Marvell brings custom silicon, optical interconnects, and silicon photonics into closer alignment with Nvidia’s ecosystem. That matters because AI systems are increasingly constrained not only by chips, but by how quickly and efficiently data can move between them.
How is Slack changing with all these new AI features?
Slack is being pushed toward a more agent-like workflow experience rather than a basic chat tool with AI layered on top. The clearest shift is reusable AI skills, which are designed to handle repeatable tasks across channels, apps, and company data. In practice, that means Slackbot could move from answering prompts to coordinating multi-step workplace actions.
What are California’s new AI vendor rules trying to accomplish?
The emphasis appears to be practical procurement control rather than broad regulation built for headlines. California is using privacy and security requirements as a condition for companies that want to sell AI tools to the state. That approach matters because procurement rules can quietly shape behavior, even when national AI policy remains uneven or unsettled.
Why would a startup like Yupp shut down after getting users and funding?
The article suggests product-market fit was the central issue, even with 1.3 million users and strong investor backing. Yupp’s model evaluation product may have had value, but not enough value to become essential as the market moved quickly. In fast AI cycles, a startup can gather attention and data, yet still lose relevance before its category fully matures.
Is Apple finally catching up in AI assistants with Siri’s multi-command feature?
It looks like a meaningful step, even if it feels overdue compared with newer AI assistants. Handling multiple requests in a single query pushes Siri closer to a chatbot-style interaction model, where users expect more flexible conversations. The larger signal is not just the feature itself, but that Apple appears to be moving with more urgency on its broader Siri overhaul.
Could energy costs slow down AI spending and data center expansion?
Yes, that is one of the clearest risks in the article. S&P Global’s warning ties massive planned AI spending to a practical limit: data centers consume enormous amounts of power, and rising energy costs can pressure returns. Even when demand for AI infrastructure is strong, electricity prices and geopolitical instability can make aggressive expansion harder to sustain.