Howmet's SpaceX Panic: Demand Validation in a Competition Costume

Howmet Aerospace shares fell roughly 8% on Monday, their worst single-session drop in more than a year, after Elon Musk said SpaceX would cast its own turbine blades at a foundry in Bastrop, Texas, so it could get gas turbines built faster for its own AI data centers. By Tuesday the stock had bounced back nearly 4%, after two major Wall Street banks argued the market had misread the situation.

The market's story was easy to follow: vertical integration is a supplier's nightmare. If SpaceX can make the hardest part itself, the logic goes, Howmet loses an order and the pricing power that goes with it. Howmet makes the precision-cast superalloy blades and vanes at the heart of jet engines and industrial gas turbines, a niche it dominates, and the scare landed on a stock that was already up more than 20% this year and had slipped roughly 13% off its record high before the news broke. Rich multiples turn into momentum unwinds quickly when a headline offers a reason.

The story leaves out who Howmet actually sells these parts to. Its industrial gas-turbine business runs to the companies that build the power plants — GE Vernova, Siemens Energy and Mitsubishi — under long-term supply agreements that extend through 2030. SpaceX is not one of those customers today. Its rocket engines are built around 3D printing rather than cast turbine blades, and gas turbines to power data centers are a new market, not an account Howmet already had on its books. The announcement touched none of the contracted revenue.

The more useful number is why Musk is bothering at all. GE Vernova ended the second quarter with 116 gigawatts of gas turbines under contract against roughly 10 gigawatts a year of production capacity; a new order today has a delivery slot in 2031. Turbine blades and vanes are some of the longest-lead-time parts in industry, quoted at 60 to 90 weeks. Musk says casting the parts himself can move a turbine's commissioning ahead by up to 18 months. When a company builds a foundry instead of standing in line, that is the market's best available evidence that the constraint is real — and that constraint is precisely the demand Howmet monetizes.

The company's own numbers were already saying it. Gas-turbine revenue grew 38% in the second quarter after rising 39% in the first and 25% across 2025. In August, Howmet raised full-year 2026 guidance to adjusted earnings per share of $5.23 to $5.31 and free cash flow of roughly $1.9 billion, with the chief executive calling the demand "extraordinary." Management has also been buying stock to back its math: $800 million through July at an average price near $248, close to where the shares trade now.

The immediate reaction on the Street was the opposite of a cut. Citi reiterated its Buy rating and placed the stock on a 30-day upside catalyst watch, and the firm argued the plan "reinforces the critical nature of HWM parts," with demand running beyond the roughly $2 billion gas-turbine revenue objective the company carries. Bernstein kept its Outperform rating and raised its price target to $328, reading the SpaceX plan as a sign of tight supply rather than a threat. A selloff that produces no estimate cuts is a multiple event, not an earnings event — the market changed what it pays for Howmet, not what Howmet earns.

The only problem is the price. Howmet trades near $255, roughly 48x the adjusted EPS it just guided to for this year — $48 for every expected dollar of 2026 profit — and about 11x sales, with a dividend yield under 0.2%. It sits about 18% below its record yet is still up 24% year to date and nearly 48% over the trailing twelve months. This is not a fallen stock at a cheap multiple. The dip is an expensive growth story at a slightly less expensive price, and "buy the dip" only works if the compounding keeps compounding.

That is why the line to watch is not the foundry in Bastrop; it is whether one foundry becomes a pattern. A single SpaceX operation coming online by 2027 is a footnote against a gas-turbine backlog the industry cannot fill this decade. The real risk to Howmet's price, and eventually its franchise, is a template rather than a plant: if hyperscalers and power developers start building their own blade capacity instead of waiting in line, demand spreads across new entrants and pricing pressure follows. Nothing announced says that is happening; it is a condition to monitor, not a reason to sell.

Musk spent real money to prove the exact thesis Howmet trades on: the bottleneck is real, and it is not going away before 2030. The panic priced a threat that was never on Howmet's books, and the buyers had the better math. The honest caveat is that better math at 48x still requires years of growth to show up — so this is a bet on execution at a premium price, with the order book, not the foundry, as the number that decides.