November 19, 2025
AI is still the dominant market narrative, but the more useful recruiting angle isn’t “AI is big.” Everyone knows that. The better angle is that AI is becoming a full capital markets, infrastructure, accounting, and M&A story.
This week’s cleanest examples are SoftBank selling its Nvidia stake, Michael Burry questioning AI depreciation assumptions, and PJM power prices surging as data-center demand strains the grid. Put together, they give you a much sharper way to talk about AI in an investment banking interview: not as a product trend, but as a financial model stress test.
AI Is Moving From Revenue Story to Cost Story
The U.S. and China are still racing for AI dominance. The U.S. leads in top-tier AI models, advanced chips, and private capital, with American investors putting about $104 billion into AI startups in the first half of 2025. China is responding with a state-led push built around scale, subsidized computing, domestic chips, and data centers.
That alone is worth knowing. But for banking interviews, the more interesting question is: who pays for the infrastructure?
AI requires massive spending on chips, servers, data centers, and electricity. That means the story flows into debt markets, equity funding, project finance, utility demand, real estate, and earnings quality. If you only talk about model performance, you’re missing the banking angle.
SoftBank is a good case study. The company sold its entire Nvidia stake, worth more than $5.8 billion, to raise cash for other artificial intelligence investments, including OpenAI, Oracle’s Stargate data centers, and onshore robot manufacturing plants. SoftBank shares fell more than 10% in Tokyo after the announcement, while Nvidia dropped 3.9% in the U.S.
Management framed the Nvidia sale as financing, not a call on whether AI is in a bubble. That distinction matters. A company can still believe in the long-term AI theme while selling a liquid winner to fund the next leg of investment. In banking terms, this is capital recycling: monetize one asset, redeploy into strategic priorities, and accept some market signaling risk along the way.
Depreciation Is Suddenly an AI Interview Topic
Michael Burry’s short view on AI gives another useful angle: accounting assumptions.
His argument is that major AI and cloud companies may be boosting earnings by understating depreciation costs. These companies are buying huge amounts of Nvidia chips and servers, but if they assume those assets last longer than they realistically will, annual depreciation expense falls. Lower expense means higher reported earnings.
Burry estimates that from 2026 through 2028, depreciation could be understated by roughly $176 billion across the industry. He also argues that companies like Oracle and Meta could have earnings overstated by more than 20% if his estimates are right.
You don’t need to take a side on the trade. For interviews, the point is that depreciation assumptions can materially affect earnings quality. If AI hardware becomes obsolete within a few years, then useful life assumptions become a real diligence issue.
If an interviewer asks about AI valuations, a strong answer might sound like this:
“The debate isn’t just whether AI demand is real. It’s whether the return on invested capital can justify the capex, and whether depreciation schedules accurately reflect how quickly AI hardware turns over.”
That’s a much better answer than saying “AI stocks are expensive.”
PJM Power Prices Show the Infrastructure Constraint
The clearest real-world constraint is electricity.
AI data centers consume enormous amounts of power, and supply can’t adjust quickly. Building new nuclear, coal, or gas capacity can take five to ten years, while wind and solar are not reliable stand-alone sources because output depends on conditions and time of day.
That supply-demand imbalance is already showing up in PJM Interconnection capacity prices. Auction prices increased from $28.92 per MW-day for 2023/2024 to $329.17 per MW-day for 2026/2027. That’s roughly an 11x jump. Household electricity bills have also been rising faster than inflation, increasing 5% to 7% year over year since the summer.
This is where a finance student can connect sectors. AI is not only a technology topic. It affects utilities, power trading, infrastructure funds, data-center developers, equipment suppliers, and potentially consumer bills. If you’re interviewing for power, utilities, infrastructure, or tech coverage, this is the bridge you want to build.
Macro Still Matters: The Fed Is Split and the Dollar Is Under Pressure
AI may be the exciting story, but rates still drive valuation.
The Federal Reserve is divided over whether to cut rates in December. Dovish officials point to weaker job growth and softer demand, with payroll gains falling from 168,000 earlier in the year to just 29,000 on average over the summer. Hawks are still worried about inflation remaining above the 2% target and the possibility that tariff-related costs get passed through next year.
The recent government shutdown made the debate harder by delaying key economic data. Private labor data also pressured the dollar, with estimates showing private employers shed 11,250 jobs per week through October 25. The WSJ Dollar Index slipped 0.2%, and the Swiss franc gained 0.6% on the day.
For valuation, the message is simple: lower rates could support multiples, but only if investors believe earnings quality and growth durability are intact. That’s why strong earnings alone haven’t been enough. Around 80% of S&P 500 companies that reported beat expectations, the highest share since 2021, but the index rose only about 1% since earnings season began. Investors had already priced in a lot.
Some large technology companies tied to the AI boom are trading at 50x or even 100x projected earnings. In that environment, a beat isn’t always a catalyst. The market wants proof that capex turns into cash flow.
Consumer and Labor Data Are Flashing Caution
There are also signs that companies are getting more conservative outside the AI complex.
Holiday hiring is cooling sharply. Job postings tied to retail, hospitality, and food service fell 16% in October compared with the same month last year. The National Retail Federation expects holiday hiring between 265,000 and 365,000, down from 442,000 in 2024 and the lowest level since the Great Recession. At the same time, job-search activity for holiday roles rose almost 27% year over year.
That’s a simple supply-demand imbalance in labor: more workers looking for seasonal jobs, fewer seasonal jobs available.
Tourism is another pressure point. Canadian travel to the U.S. has declined, contributing to an estimated $5.7 billion loss in international tourism spending in 2025 compared with last year. October return trips among Canadians traveling to the U.S. fell 24% by air and 30% by land. Since Canadians accounted for 28% of international visitors to the U.S. in 2024, that drop matters for hotels, restaurants, retailers, and border-state local economies.
Corporate Strategy: Toyota, Burger King China, and Airlines
Several company-specific stories are useful for interviews because they show how management teams are repositioning around demand, risk, and local execution.
Toyota began production at its new $13.9 billion battery plant in North Carolina and confirmed plans to invest another $10 billion in the U.S. over the next five years. The timing is interesting because EV demand has weakened while hybrid demand remains strong. Toyota has 51% share of the U.S. hybrid market through the third quarter of 2025, and U.S. sales rose 9.9% to more than 1.3 million vehicles.
That’s a strategic flexibility story. Toyota is investing in batteries, but its near-term advantage is hybrid scale.
Burger King China is another example. CPE agreed to spend $350 million for an 83% stake in Burger King China, while Restaurant Brands International keeps a 17% stake and a board seat. The goal is to grow from the current footprint to about 2,500 sites within five years and more than 4,000 outlets within ten years. The deal supports RBI’s broader asset-light approach while leaning on a local partner for execution in a competitive consumer market.
Airlines faced a different issue: operational risk from the government shutdown. Federal officials ordered a 10% cut in domestic flights across 40 major airports during the 41-day shutdown, citing air-traffic safety. Airlines questioned the data behind the order, but the FAA treated the cuts as non-negotiable. More than 5 million passengers were affected.
For banking candidates, this is a reminder that operational risk can quickly become revenue risk, customer risk, and regulatory risk.
M&A Is Still Active Where the Thesis Is Clear
Despite macro uncertainty and higher rates, deals are still happening in sectors with specific tailwinds.
KKR agreed to sell Novaria Group to Arcline Investment Management for $2.2 billion in cash. Novaria supplies precision components and specialty manufacturing to aerospace and defense customers. Under KKR’s ownership, the company completed 13 add-on acquisitions and more than tripled in size.
This is classic private equity value creation: buy a platform, complete add-ons, scale the business, and exit into a market where buyers still value structural growth. Aerospace and defense also benefit from commercial aircraft demand and higher geopolitical security budgets.
Real estate M&A also had a major example. SBB agreed to sell a $3.4 billion portfolio of elderly care homes to Public Property Invest ASA. SBB plans to use roughly $1.6 billion of proceeds to repay bonds and loans. PPI is buying more than 700 properties through cash and bridge financing and will issue 446.9 million new shares to SBB at 23 kroner per share. After the transaction, PPI’s portfolio will reach 841 properties and roughly 53 billion NOK, with 50% of assets concentrated in elderly care and healthcare.
That deal is a balance sheet repair story for SBB and a scale-building story for PPI.
How I’d Use This in Interviews
If you want one polished market view, use this:
“AI remains the dominant growth theme, but the market is starting to underwrite the costs more carefully. SoftBank’s Nvidia sale shows that even AI bulls need liquidity to fund the next wave of capex. Burry’s depreciation argument raises questions about earnings quality. PJM’s 11x capacity price increase shows that power supply is a real bottleneck. So I’d frame AI less as a simple revenue-growth story and more as a capital intensity and infrastructure story.”
That answer gives you several banker-friendly discussion points: capital allocation, accounting, infrastructure, utilities, valuation, and financing. It also avoids the lazy “AI bubble or not?” binary.
That’s the standard you should aim for in recruiting. Don’t just memorize headlines. Turn them into a thesis, connect the sectors, and explain why a banker would care.