Palantir, Anthropic and the factory behind the AI boom

• Anthropic’s latest deals show AI value shifting from flashy models toward trusted deployment and scarce compute capacity.

• Palantir stands out because large institutions value secure workflows and trusted access as much as raw model performance.

• ASML and TSMC earnings are the real test of whether AI demand is turning into sustained industrial spending.

Artificial intelligence still looks like a software story from the outside. You see a chatbot, a coding tool, or a clever demo and assume the magic lives in the screen. This week’s Anthropic news suggests the opposite. The real battle is moving underneath the interface, into enterprise data, operational workflows, and the chip infrastructure required to keep these models useful at scale.

The model is not the product anymore

Anthropic helped re-open one of Wall Street’s sorest questions on 9 April 2026, when its new Claude Mythos model was powerful enough to unsettle software investors and was being previewed only to a small group of major partners for defensive cybersecurity work. That same week, Anthropic expanded compute partnerships with Google and Broadcom, struck a new cloud-capacity deal with CoreWeave, and it is exploring whether to design its own chips. That is not the behaviour of a company worried about demand. It is the behaviour of a company worried about supply.

That matters for investors because it changes the question. The old question was which model wins. The new one is who controls the route from raw intelligence to useful work. In other words, who has the data, who has the workflow, who has the trust layer, and who can still get enough compute when everyone else is queuing outside the data centre.

The trust layer gets political

If Anthropic represents the raw power of the new model cycle, Palantir shows what happens when that power collides with the real world. Palantir builds software that helps governments and companies turn data into decisions and actions. That may sound less glamorous than the latest chatbot demo, but it is often where artificial intelligence becomes useful, accountable and expensive enough to matter. In March, the Pentagon planned to adopt Palantir’s Maven Smart System as a formal programme of record, a sign that Palantir is moving deeper into the operating layer of defence artificial intelligence rather than sitting at the edge as a tools vendor.

The wrinkle is Anthropic. Anthropic’s Claude model was used inside parts of Maven, just as Anthropic fell into a bitter dispute with the Pentagon over defence guardrails and was labelled a supply-chain risk. Anthropic is still talking to the Trump administration about its latest Mythos model despite those tensions. That makes the Palantir-Anthropic story useful for investors because it captures the next phase of artificial intelligence in one neat, slightly awkward package. The models are getting stronger, but access to them can still be restricted by politics, procurement rules and national-security concerns.

That leaves Palantir in an interesting position. Its value is not that it owns the best frontier model. Its value is that it sits closer to the mission, the workflow and the customer relationship. In simple terms, models may come and go, but the company that helps an institution run them safely inside a real operation can still keep its seat at the table. That is why this part of the AI story matters. The market often talks as if intelligence alone wins. Defence and enterprise buyers are quietly voting for something more boring and more durable: trusted deployment.

The real receipts arrive this week

If Anthropic is the demand signal, ASML and TSMC are the industrial receipts. ASML makes the lithography tools used to print advanced chips. TSMC manufactures many of the world’s most advanced processors. If they sound calm and confident this week, investors can reasonably conclude that artificial intelligence spending is still moving from slide deck to factory floor. If they sound cautious, the market will hear it immediately.

ASML reports on 15 April 2026. In January, the company reported record quarterly bookings of EUR 13.2 billion and a raised 2026 sales outlook of EUR 34 billion to EUR 39 billion, driven by artificial intelligence demand. Since then, investors have had a fresh reason to worry after new restrictions on exports to China could hit sales of deep ultraviolet lithography tools. So this quarter is not just about demand. It is about whether demand is strong enough to outrun politics.  

TSMC reports on 16 April 2026. The company’s own guidance for the first quarter points to revenue of USD 34.6 billion to USD 35.8 billion and gross margin of 63% to 65%. Bloomberg analysts expect a fourth straight quarter of record profit, driven by demand for 3-nanometre chips and advanced packaging that still exceeds capacity. That is the key point. If Anthropic and its peers need more compute, TSMC is where that ambition becomes a physical bottleneck. 

The cracks to watch

The risks are not mysterious. First, enterprise adoption can still disappoint. It is one thing to sign a partnership. It is another to turn that into large, sticky spending.

Second, policy risk is no sideshow. The Anthropic Pentagon dispute shows that model access can become political overnight, and ASML’s China exposure remains vulnerable to export controls. A strong product cycle is helpful. A minister with a pen is still unhelpfully powerful.

Third, valuations leave little room for a dull quarter. When expectations are high, even “good” can trade like “not good enough”.

Investor playbook

• Watch the stack, not just the model. Data, workflow and chip capacity usually outlast the latest demo.

• Treat ASML and TSMC as demand detectors for the whole artificial intelligence chain.

• Look for proof of adoption, not partnership headlines alone.

• Expect sharp sentiment swings. AI enthusiasm still has the emotional stability of a shopping trolley on a hill.

ASML and TSMC matter because they tell us whether that work is backed by real industrial spending. So the next time the market gets distracted by the latest clever chatbot, it is worth remembering where the durable value often sits. Not in the demo, but in the data pipes, the workflow engine and the very expensive machines humming behind the curtain.

This material is marketing content and should not be regarded as investment advice. Trading financial instruments carries risks and historic performance is not a guarantee of future results.

The instrument(s) referenced in this content may be issued by a partner, from whom Saxo receives promotional fees, payment or retrocessions. While Saxo may receive compensation from these partnerships, all content is created with the aim of providing clients with valuable information and options.