AI Market Outlook After a Rough November Selloff

Investors discussing the AI market outlook at a conference

The AI investment trade had a brutal November, and the mood at the TomorrowX conference in Austin showed it. Investors, independent writers and tech executives gathered in what one attendee called the reddest week in recent memory, trying to answer a simple question: is this a buying opportunity or the start of something worse?

Data center buildouts met pushback after the midterms. Nvidia’s leadership turmoil, including Jensen Huang’s departure and talk of a possible return, dominated headlines. A Hunterbrook short report on AMD over China shipments added to the pressure, and AI stocks sold off hard.

Gloom over the AI market outlook

Sentiment in Austin was dark. Huang’s exit from Nvidia hit some attendees personally. Others had watched their holdings fall sharply, in some cases below zero, and many investors had rotated out of AI chips and infrastructure into other sectors.

The open question was whether capital spending would slow. SemiAnalysis recently cut its projection for AI capex by about 40%, from roughly $1.5 trillion to about $900 billion by 2030. Such sharp revisions are hard to trust, since they can swing back up just as fast.

The AI market outlook stayed murky. Investor Stanley Druckenmiller, CEO of Duquesne Family Office, offered no confident call. He said he has never had less clarity about where things will stand in twelve months, and his answer is to stay out of the game. His version of the old rule: when he lacks clarity, he doesn’t play.

An AI market outlook built on early innings

Optimists still call it early innings, and math is their evidence. The Riemann hypothesis stood unsolved for more than 150 years, and two solutions appeared on the same day. Less than two months earlier, OpenAI reported a solution to the Navier-Stokes existence and smoothness problem. Mathematicians are increasingly becoming verifiers of machine-generated proofs.

Druckenmiller said he expects cancer to be a solved problem within ten years. That is a prediction, not a result, but the bull case treats AI as a new tool, the way PCR was for genomics. Chatbots and computer-based knowledge work are only the first uses. The comparison is to railroads, which enabled big-box retail, and the retailers earned better returns than the rail infrastructure did. The true beneficiary of AI hasn’t been identified yet.

This view doesn’t depend on AGI. Useful intelligence can come from specialized models for biology, space research or chip design, built by large companies with domain expertise. Those could be multi-trillion-parameter systems, which would be even better for compute demand than a single general intelligence.

Where infrastructure investors are looking

Hunting for bottlenecks is finished as an investing strategy. Order backlogs stretching into 2028 and 2029 mean little if supply can’t be brought online to turn them into revenue. Areas worth watching:

  • TSMC, which has kept capital spending conservative and so may avoid a chip glut. TSMC fits the profile of a company with a durable moat, steady demand and solid execution. Intel has none of those today.
  • Wafer fab equipment, buoyed by Elon Musk’s TeraFab plans. Musk said at the event that the first test chip arrives in 2027 and phase-one volume production by the end of 2028. Rumors put the target at 1 million logic wafers and 700,000 memory wafers a month by 2032. None of this has materialized yet.
  • Power efficiency across the AI stack, from GPUs and stacked memory to co-packaged optics and better quantization.

The efficiency argument rests on arithmetic. A human brain uses about 20 watts, so 8 billion people represent roughly 160 GW of biological compute. Projected AI data center demand is more than twice that by 2030. A million GPUs at 2-3 kW each is brute force. Micron CTO Scott DeBoer said at the event that vertically stacked memory on logic is the company’s top priority, because many customers are asking for it.

Tasks instead of tokens

Druckenmiller said he is looking for AI use cases rather than infrastructure. Infrastructure may work as a tactical trade because positions are washed out, he said, but he doubts it will drive returns the way it did over the past four years. His most interesting use case is an AI lab itself. In his view, a lab with the top intelligence in the world could stop selling tokens and start selling tasks.

Semiconductor stocks should keep paying off, though not at the frantic pace of recent years. In the near term, the advice is to look for chip companies that fell further than they deserved. The next big winners are likely to come from outside semis.

Source: viksnewsletter.com

Written by

Saksham Bhargava

Saksham Bhargava runs Alteroid and has been writing about consumer technology for over five years, covering iPhone, Android, Windows, AI tools and gaming. With an MCA and six years of hands-on web development experience, he tests the features he writes about on his own devices and focuses on clear, practical guides that solve real problems. Away from the screen, he enjoys travelling to places most tourists skip.