Ban Anthropic is now a fashion
Anthropic may have developer love, but unfortunately, that matters less when governments, security teams, and platform owners make the final call.
Bending Spoons IPO = the end of Saas?
Ever heard of Bending Spoons? It’s the Italian software group behind acquisitions such as Evernote, WeTransfer, Vimeo, Eventbrite, Brightcove, and AOL, which raised about $1.68 billion in its IPO after pricing above its range. The company doesn’t follow the typical PE playbook, but turns a sunsetting Saas into a durable cash machine.
→ FT
Bending Spoons is never “innovative” in the usual Silicon Valley sense. But treating software like any asset. If AI can really compress the cost of maintaining old products, then the graveyard of mature SaaS may become more valuable, not less.
I’m going through their IPO documents. Stay tuned.
SK Hynix lists and memory stock tight
SK Hynix is launching a $28 billion Nasdaq listing via depository receipts, riding an AI-memory-driven surge that’s pushed its stock up ~273% this year despite recent volatility.
GPUs have had a narrative premium for a while now, while memory is the next star of the show. AI does not just need more chips, but more bandwidth, more packaging, more power, more everything. SK Hynix is a reminder that the bottleneck is often one layer below the thing everyone is talking about. Same as Micron.
So in my Micron’s earnings report analysis, you’ll find the answer to the following:
Why does it suddenly get so much attention?
Why does the stock price boom?
The RAM shortage and why does AI make this worse?
How! How this memory party is not a good sign for companies adopting AI, and even worse news for AI labs?
And finally, what to look out for?
Is Meta giving up on its AI and starting to sell compute?
Meta Platforms is launching Meta Compute, which sells its excess AI computing capacity to enterprise customers, putting it in direct competition with the likes of AWS.
Meta renting compute is a strange reversal.
The company that depends on cloud capacity now suddenly becomes a competitor to the same suppliers. AI infrastructure is starting to look less like a clean stack and more like everyone selling to, buying from, and undercutting everyone else.
There’s also an upside to telling a good story in an investor relations announcement. So the excess compute is no longer a cost center, but a product.
From “we overbuilt” to “we are launching a cloud business,” what an impressive U-turn!
OpenAI’s Public Stake
Rumor has it that OpenAI is proposing that the U.S. government take a 5% stake in the company, possibly as part of a broader public-benefit fund for major AI labs.
As I mentioned in my Anthropic export ban warning piece, this isn’t going to be a one-off. So it only makes sense that OpenAI and absolutely other AI labs are searching for ways to be on the ‘government’s’ side (whatever that means, and in whatever means).
For nearly a decade, Tech companies asked governments to stay out of the way. Now they are offering a piece of the pie for the government’s involvement.
We should be worried. Because if AI becomes national infrastructure, “using AI” stops being a product decision and becomes a potentially political dependency embedded in your workflow.
AI Makes Shampoo
L’Oreal’s been using AI to basically speed up how they invent new products, like they took some skincare molecule and turned it into this new collagen shampoo, four times faster than usual. Or Mondelez is doing the same thing with AI-generated recipes, which generated gluten-free Golden Oreos.
→ Reuters
Everyone is now using or being impacted by AI. Even those who've never heard of it (and never will) are still under its influence.
As I said two years back, the only way for AI to succeed is if it becomes something camouflaged in the background:
Like electricity and the internet before it, AI WILL (at some point) transform our world in ways beyond everyone’s imagination. But history suggests a few key takeaways:
Expect a shakeout: Most AI startups today won’t survive long-term. That’s not pessimism; it’s pattern recognition.
Beware of inflated promises: If it sounds too good to be true in tech, it usually is.
Think long-term: The real impact of AI will likely unfold over decades, not quarters.
Focus on value: Look for products that improve efficiency, accuracy, or user experience. High value and high impact on users are the best predictors of long-term success.
Claude gets banned, this time, not by the US.
Alibaba has reportedly banned employees from using Anthropic’s Claude Code and told them to remove Claude models from work computers.
→ Reuters
Deglobalization 2.0
As the US government didn’t want others to have access to Anthropic, apparently, the biggest Chinese software enterprise agrees. Alibaba also doesn’t want its staff to have access to Claude.
I suspect, though, this is less political but everything commercial:
to avoid code leakage, a coding assistant may touch proprietary code, internal architecture, deployment scripts, bug reports, credentials, or product plans…
to avoid customer and business data exposure, this is quite self-explanatory
dependency on a U.S. vendor, can’t blame them, the Fable 5 export ban, as I said, a huge alarm to many.
to train and use their own models ← like Qwen or Wan
If you were in Alibaba's shoes, would you allow your staff to use others’ models?
Microsoft picks Copilot over Claude code
Microsoft is canceling most internal Claude Code licenses and moving employees toward GitHub Copilot CLI, its own coding assistant. Claude Code had reportedly become popular inside Microsoft, but the company wants tighter control, lower costs, and deeper integration with its own developer platform.
→ Forbes
The argument is exactly the same as why Alibaba banned its staff from using Claude.