Closing Bell Overtime: 10/15/26

Amazon.com

Across the shows: 9 speakers in other episodes, 5 bullish and 3 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBullish

  Amazon is among the full-stack hyperscalers where AI economics will ultimately settle.

  Full-stack hyperscalers like Microsoft, Google and Amazon have the raw chips, services and software needed to support AI, while neoclouds lack that breadth. Niles argues the AI boom's economics will settle at these hyperscalers, which are the ones able to capture value from excess compute and services.

  “Microsoft, Google, and Amazon because not only do they have the raw chips, they have the services, they have the software, they have everything to support that.”at 5:37

CoreWeave

Across the shows: 18 speakers in other episodes, 3 bullish and 18 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBearish

  CoreWeave's neocloud model faces a wave of bankruptcies as hyperscalers' excess capacity contracts unwind.

  Hyperscalers like Microsoft and Google are contracting out excess capacity to neoclouds such as CoreWeave rather than overbuilding their own. AI is seen as the most transformational technology since the internet, so overinvestment is inevitable. Neoclouds sit at the layer where that overbuilt capacity lands, so Niles expects a huge wave of bankruptcies when the cycle implodes. He agrees that neoclouds may not have a place in the ecosystem once compute is rationalized.

  “Where you're gonna have problems, and I've been saying this for I don't know how long, is at the neo cloud level. Because what the Microsofts and the Googles are doing is they're contracting out with the likes of CoreWeave for that excess capacity.”at 5:48

Microsoft

Across the shows: 10 speakers in other episodes, 7 bullish and 2 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBullish

  Microsoft will ultimately win enterprise AI because customers trust its safe Microsoft 365 AI access.

  Microsoft has taken advantage of fears that AI will displace jobs, and corporations feel safe with its Microsoft 365 license, which gives them AI access in a controlled environment. Microsoft has also done a good job with the Copilot overhaul. Together these position it as the ultimate enterprise winner in the AI layer.

  “...corporations feel safe because they've got a Microsoft three sixty five license... They've done the Copilot overhaul. I think Microsoft ultimately wins in enterprise.”at 7:12

Alphabet

Across the shows: 13 speakers in other episodes, 14 bullish and 3 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBullish

  Google, via the Android ecosystem, will be a consumer AI winner regardless of which model runs underneath.

  Consumers will talk to their phones, and the Android ecosystem controlled by Google will be the interface. Consumers won't care which underlying model runs, because Google and Apple will route each query to whatever model is best and cheapest for the job. That gives Google the consumer end of the AI stack.

  “...it'll be ultimately be the Android ecosystem, which is obviously Google, and it will be Apple... those two companies will route your query to whatever is best to get the job done at the best price.”at 7:38

Apple

Across the shows: 21 speakers in other episodes, 9 bullish and 7 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBullish

  Apple will be a consumer AI winner, routing queries to the best model via its iPhone ecosystem.

  Consumers will talk to their iPhone or Android device and will not care which model runs underneath. Apple and Google will route each query to whatever model gets the job done at the best price, so the device platform captures the consumer AI value rather than the individual model makers.

  “...it'll be ultimately be the Android ecosystem, which is obviously Google, and it will be Apple... you'll either talk to your iPhone or your Samsung or whatever it is running Android.”at 7:38

Anthropic

Across the shows: 38 speakers in other episodes, 18 bullish and 39 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBullish

  Anthropic should make it through the shakeout because its enterprise focus means businesses pay.

  Anthropic has focused on the enterprise market from day one, and businesses will pay for AI, unlike consumers who expect free products. That makes it better positioned than OpenAI, which is shifting to enterprise late. Niles thinks Anthropic will get through the coming shakeout okay, though his view is more relative than enthusiastic, and its IPO has also been delayed.

  “Anthropic from day one has been focused on the enterprise market. Why? Because businesses will pay.”at 9:31

OpenAI

Across the shows: 31 speakers in other episodes, 17 bullish and 26 bearish calls, past 30 days

• Dan Niles · Niles Investment ManagementBearish

  OpenAI is at the epicenter of LLM-layer risk; Niles has real concerns about it.

  Niles has repeatedly said OpenAI is at the epicenter of LLM-layer problems. OpenAI's consumer-first model is weak because consumers expect free products, which forced it to pivot to enterprise this year. Its IPO was supposed to come first but has been pushed to next year. Deals have also been pulled, including an Nvidia-backed Australian deal and AirTrunk's Singapore IPO, which he links to the risk of a 2000-style funding squeeze.

  “But I have, you know, real concerns about OpenAI. And let's not forget, both of these IPOs have gotten pushed out, right? Anthropic was supposed to come earlier.”at 9:45

AI chips & semiconductors

• Dan Niles · Niles Investment Management

  Long-term supply agreements for chips are worthless, as customers renegotiate or exit once demand falls.

  Long-term agreements struck during COVID proved worthless because customers went back and demanded changed terms. Microchip enforced its contracts and suffered four straight quarters of revenue down over 40% before its CEO canceled them. Enforcing contracts means customers like Apple will take their business to competitors when the contract expires. Niles says the current AI chip LTAs are just as worthless.

  “Those long term contracts are worthless in my mind. I mean, go back to what happened during COVID. If you remember, there were a lot of long term agreements struck during COVID.”at 11:24

• Eric Johnston · Cantor Fitzgerald Asset Management

  Semiconductors look attractive and should bounce in the next couple of weeks.

  Johnson closed a bullish tech position after a big run, but after a sharp de-risking and very high fear, he sees semiconductors as the attractive part of tech. Near-term catalysts include hyperscaler earnings showing positive capex, possible news around Anthropic going public, and AI worries fading after the midterms. Semis also have strong balance sheets.

  “I actually think that within tech, I think semis do look attractive, and I think they are going to get a bounce from here.”at 20:58

• John Kolovos · Macro Risk Advisors

  Semiconductors are lagging the Magnificent Seven since June's speculative peak; oversold ratio risks forced liquidation.

  Semiconductors are higher-beta names, so their underperformance is a warning sign. Kolovos uses the ratio of the semiconductor index to the Magnificent Seven, which peaked in June as the speculative peak in technology and has kept falling. The ratio is now testing its 200-day moving average and is oversold on RSI. If the trend continues, he expects forced liquidation across many of those names. A rebound in CoreWeave is encouraging, but the trend needs to reverse.

  “What we need to be mindful with the semi trade is that this is more of the higher beta part of the market. ... this ratio peaked back in June. ... What I'm afraid of here is that this trend were to continue, it's gonna create forced liquidation and a lot more of those names.”at 42:09

AI infrastructure & data centers

• Seth Meyer · Janus Henderson Investors

  Hyperscaler capex will hold near $1.2 trillion next year despite CoreWeave's credit stress.

  Hyperscalers with fortress balance sheets will keep spending regardless of CoreWeave's credit trouble. CoreWeave's roughly $50 billion of spending matters only at the margin. The economic impact on GDP from a build-out slowdown is likely small. The real constraints are power, local policy and permits, people, and the paper to finance projects. Investment-grade issuers can still fund, though high-yield financing is harder.

  “the CapEx budgets of the largest hyperscalers, right, those with the fortress balance sheets, this means very little. We're still gonna see $1,200,000,000,000 in spend next year.”at 16:00

• Brandon Gomez · CNBC

  Each megawatt of data center capacity needs about 30 tons of aluminum, driving global aluminum demand.

  Gomez says every megawatt of data center capacity built requires about 30 tons of aluminum, so a gigawatt needs about 30,000 metric tons. He argues this data center build-out is one of the drivers of the strength and demand for aluminum seen globally.

  “For every megawatt of data center capacity that's built, you need about 30 tons of aluminum. So, if you think about a gigawatt, you need about 30,000 metric tons of aluminum.”at 35:45

US Treasuries & yields

• Eric Johnston · Cantor Fitzgerald Asset Management

  Long Treasury yields have a ceiling, and the 3%-plus term premium is attractive to buyers.

  The 10-year yield is only about 80 basis points off its three-year average, which is not very meaningful for an economy driven by AI capex. Yields should level off because Treasury will act to control them. Inflation breakevens are contained at 5.35% over five years despite higher oil. The term premium is over 3% for the 10-year and 30-year above inflation, which attracts demand.

  “But I think that yields do sort of level off here, and that's been our view that there is a ceiling to them. ... It's over 3% for the ten year, over 3% for the thirty year above inflation.”at 22:48

• John Kolovos · Macro Risk Advisors

  The 10-year yield's pullback is consolidation, not reversal; it should fall toward 5% before rising again.

  The 10-year was projected to peak around 5.15% and has since pulled back, with bearish sentiment, COT data, RSI and some Elliott wave analysis cited. Yesterday's move looked like a selling climax after CPI, and the follow-through decline looks to Kolovos more like a flight to safety than a true reversal in trend or risk sentiment. He expects the consolidation to carry the yield down to about 5% before it moves higher again.

  “I'm interpreting that more as a flight to safety than a proper reversal in trend and also in risk sentiment. But for now, I think we have enough of a decline in consolidation to see it work its way down to about 5% before we work our way higher once more.”at 41:38

US stock market outlook

• Eric Johnston · Cantor Fitzgerald Asset Management

  Heavy AI-trade concentration is a major market risk that will cap valuation multiples.

  Market gains and the economy depend on one AI trade. Even retail is propped up by the wealth effect, since consumers have their highest-ever equity exposure after equities rose about 20% a year for four years. If the AI trade rolled over, the slowdown could cascade across the economy, so investors will keep a cap on the multiple. Johnson calls this the world investors will be in for a while.

  “I think it's the world that we're gonna be in, and I think that remains a major risk for the market and for the economy, the fact that the concentration is so heavy.”at 24:40

• John Kolovos · Macro Risk Advisors

  S&P 500 breadth is deteriorating and nearing a first sell signal since 2025, raising downside risk.

  About 30% of the market is below its 50-day moving average and almost 40% is below its 200-day. The index is about to get a sell signal for the first time in eighteen months. The last sell signals came before the 2025 and 2022 bear markets. Kolovos says this configuration must be taken seriously, as it could precede a deeper decline. A sell signal means volatility and left-tail risk, meaning the probability of a market mistake, are rising.

  “The thing that concerns me the most here right now is that the internal trends of the market are starting to deteriorate quite significantly. ... for the first time in eighteen months, we're about ready to get a sell signal with 30% of the market underneath its fifty day moving average.”at 39:45

Credit markets

• Seth Meyer · Janus Henderson Investors

  AI debt repricing is healthy; spillover into broader credit has been small and name-specific so far.

  Credit markets are repricing AI infrastructure risk, which Meyer sees as healthy after a period in which capital was cheap. Investors now distinguish cost of capital between companies, with Oracle and Broadcom bonds 15 to 20 basis points wider. The high-yield market is only about 50 basis points wider, which is not meaningful given how tight spreads were. Weakness has been idiosyncratic and has not yet hurt power or construction issuers.

  “So we haven't yet seen it spill over to where, you know, a power company is having a hard time getting financed. Construction companies would have a hard time getting financing. We haven't seen that yet. It's been very name specific and idiosyncratic.”at 18:18

Middle East & Iran

• Brandon Gomez · CNBC

  About 40% of Hormuz-region aluminum smelting went offline; recovery is slow and highly uncertain.

  When the conflict started, around 40% of aluminum smelting capacity within the Strait of Hormuz was taken offline. That capacity is slowly returning to the market, but Gomez says there is a massive amount of uncertainty over how quickly it can come back. This supply loss is part of what is shaping aluminum pricing.

  “When the conflict started, around 40% of the capacity of aluminum smelting production within the Strait Of Hormuz was taken offline. That is slowly coming back to market today. But, as you can imagine, there's a massive amount of uncertainty of how quickly that can come back.”at 37:16

Copper & industrial metals

• Brandon Gomez · CNBC

  Aluminum demand is resilient and should keep growing into 2027, except in construction, which depends on rates.

  In 2026 every major North American end market grew in demand except building and construction, and Europe showed the same pattern. Gomez expects continued demand growth for primary aluminum into 2027, again with the exception of building and construction. He says construction demand will depend on where interest rates go over the next year.

  “We see strong resilient demand, in both of our end markets. In 2026 in North America, every one of our major end markets grew from a demand perspective with the exception of building and construction. ... we see continued growth on the demand side for primary aluminum with the exception of building construction, and that'll really depend on where interest rates go over the next year.”at 37:38

US elections & politics

• Pippa Stevens · CNBC

  Data center opposition is becoming a bipartisan midterm issue that will drive turnout in Ohio races.

  Local opposition to Advantage Data Centers' 500 MW project in Oregon, Ohio has grown in recent months. A grassroots group gathered enough signatures for a ballot measure to cap data centers in the area at 25 MW. Candidates in Ohio's 9th District and at the Senate level have attacked opponents over their data center stances, after previously avoiding the issue. Labor and teachers' unions back the project, so the issue cuts across party lines. Stevens says the opposition or support will push turnout up next month.

  “And there is no doubt here that data center opposition or support is going to lead to much higher turnout at the polls next month.”at 45:20