Samsung’s $80 Billion Quarter Meets the AI Bubble Test — Daily Tech and AI, October 8, 2026

AI data center with memory chip modules in the foreground, illustrating the AI memory boom behind Samsung's record quarter.

Record memory-chip profits collide with a shaky neocloud IPO — plus a cheaper Claude, Tencent’s $5 billion bond plan, Denmark’s deepfake ban, and Nvidia-powered AI PCs.

Samsung Posts the Biggest Quarterly Profit in Tech History

On October 8, Samsung said it expects an operating profit of 107.4 trillion won — roughly $80 billion — for the July-to-September quarter. That is a nearly nine-fold jump from a year earlier, and it came in ahead of what analysts expected.

It is also a world first: no technology company has ever topped 100 trillion won of profit in a single quarter, and this is Samsung’s fourth straight record quarter. The engine behind it is simple — memory chips. Booming demand from artificial intelligence data centers has created a deep global shortage of DRAM, NAND, and the high-bandwidth memory that sits next to AI chips, and chip prices have soared as a result.

Samsung and Micron expect the shortage to last into 2027 and possibly into 2028. But here is the telling detail: Samsung’s stock barely moved on the news, and shares are still down about 25 percent from their record high in June. Investors are asking whether this extraordinary earnings run can survive once the supply crunch eventually eases and memory prices stop climbing.

A Shaky Neocloud IPO Pricks the AI Bubble

That worry is exactly why the next story matters. On the same morning Samsung announced record profits, Reuters Breakingviews flagged trouble at Firmus Technologies, a Nvidia- and Blackstone-backed data center operator that is supposed to start trading later this month. The company has reportedly cut its IPO price from as high as A$11 a share to around A$8.25, shrinking the raise from A$7.9 billion to about A$5.9 billion and cutting its equity value from A$44 billion to A$33 billion. Trading starts October 23.

So you have record profits and a struggling IPO on the same morning. That is the AI buildout in a nutshell right now: real demand and real money on the infrastructure side, but investors are getting pickier about what they will pay for the next wave of capacity. If the money gets choosier, the boom starts to look a lot more like a normal business cycle.

Tencent and SoftBank Tap the Debt Markets

Speaking of spending money, Tencent is looking to borrow a lot of it. Bloomberg reports that the Chinese tech giant is considering raising up to $5 billion through an offshore bond sale, possibly as early as this month, to fund its AI and computing infrastructure plans. This would follow a $4.66 billion bond offering in June, its biggest debt deal since 2020.

Tencent is not alone in tapping the debt markets. Last month, SoftBank raised about $11.1 billion through a junk bond offering to fund its AI plans. The money spigot for AI infrastructure is still wide open, but analysts keep warning about stability — because this borrowing binge is happening while investors are already questioning how long the boom lasts and whether all this spending ever earns a return.

Anthropic Fires a Shot in the AI Price War

Anthropic launched Claude Haiku 5.5 on October 7, the third model in its Claude 5.5 family in about a month — and it is a clear shot in the AI price war. For prompts under 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens: a roughly 90 percent price cut on the small end, matching what OpenAI charges for its GPT-6 Luna model.

Anthropic says average running costs are about 75 percent lower than the previous Haiku, even after accounting for a new tokenizer that counts more tokens per word. The model also carries a 1 million-token context window and an adjustable effort setting — a first for the Haiku line — so developers can dial between cheaper and smarter on the same request. It is also the first Haiku with safeguards for a narrow set of high-risk cybersecurity requests.

This is the model tier that powers the boring but expensive stuff: classification, summarization, extraction, live support, voice agents, and subagents doing work for bigger models. A price war here is exactly what enterprise AI teams have been waiting for, because these small models get called millions of times a day. And the timing is not an accident — Anthropic is expanding its lineup ahead of a planned IPO, and cheaper models mean bigger volumes and more developer lock-in.

Denmark Moves to Ban Nonconsensual AI Deepfakes

In Europe, Denmark’s culture minister Zenia Stampe will propose a bill that would ban the digital copying and sharing of lifelike representations of a person — including their appearance and voice — in a crackdown on AI-generated deepfakes.

The bill targets what Stampe says is a growing problem: manipulated videos and images that spread misinformation and make people doubt what is real. One of the main goals is to force tech giants to take down manipulated content quickly so it causes as little harm as possible, and it would protect artists whose performances and work could be imitated and shared. There is an exception for caricatures, parody, and satire that criticizes society and people in power.

The proposal still needs parliamentary approval, and Stampe hopes it takes effect around the turn of the year. This comes just days after Norway proposed a temporary ban on AI glasses in public places. Scandinavia is quietly becoming the testing ground for Europe’s AI rules, and the rest of the continent will be watching both bills closely.

Microsoft and Nvidia Reinvent the PC

Microsoft revealed specs and pricing on October 7 for the Surface Laptop Ultra, its new PC built on Nvidia’s RTX Spark chip. The lineup breaks down like this:

  • Surface Laptop Ultra: starts at $2,600, tops out around $5,900. Runs AI models with more than 120 billion parameters right on the device — for free, without sending anything to the cloud.
  • Surface RTX Spark Dev Box: a $6,000 machine aimed at developers.
  • Windows 11 Execution Containers: a new feature that sandboxes AI agents, so they can act on your behalf inside a controlled boundary instead of having free rein over your machine. It will roll out to all Windows 11 users, not just people who buy the new hardware.

Microsoft CEO Satya Nadella framed it as the next era of the PC: having a model is not enough — you need an orchestrated layer combining models, context, memory, and the ability to act. Nvidia CEO Jensen Huang put it more dramatically, saying they have reinvented the computer as we know it. Microsoft is even offering up to $1,000 cash back to MacBook Pro owners who trade in. They want this one badly.

Quick Hits: Meta’s Muse Comes to the iPad, Mecka AI Raises $60M

Meta’s Muse keeps expanding. The company launched a dedicated iPad app on October 7, just a month after Muse debuted on phones — a milestone it took Instagram fifteen years to reach. The assistant has passed 6.6 million installs since its September 8 launch, according to Sensor Tower, and Meta is adding business connectors like Asana, Figma, QuickBooks, and Zoom, plus retail partners including Best Buy, Walmart, and Wayfair that let the agent shop and complete transactions for you.

And Mecka AI, a Toronto startup founded in 2024, raised a $60 million Series B led by Sequoia, with Nvidia and Microsoft’s venture fund M12 joining in. TechCrunch reports the round values the company at around $500 million. Mecka pays people to record themselves doing everyday tasks like making coffee or fixing cars, wearing body sensors and using smartphones, and turns that motion data into training sets for humanoid robots. The pitch: do for robotics what Scale AI did for large language models.

Sources

  • Samsung’s record Q3 profit forecast on the AI memory boom: Reuters
  • Firmus neocloud IPO price cut pricks the AI bubble: Reuters Breakingviews
  • Anthropic launches Claude Haiku 5.5: Reuters
  • Tencent mulls $5 billion bond sale for AI ambitions: Reuters
  • Denmark plans ban on sharing AI deepfakes without consent: Reuters
  • Microsoft’s Nvidia-powered Surface AI PCs: TechCrunch
  • Meta’s Muse launches on iPad: TechCrunch
  • Mecka AI’s $60M Series B for robot training data: TechCrunch

Adapted from the October 8, 2026 episode of Daily Tech and AI.

Daily Tech and AI is AI-voiced and produced by Angelo Ruiz.

Angelo Ruiz Avatar

More about Angelo · LinkedIn

Leave a Reply

Your email address will not be published. Required fields are marked *