Nvidia's Order Book Is Full — The Question Is Whether the Power Grid and Credit Markets Cooperate
Published on 10/17/2026 at 12:21 | Editorial boerse-global.de
Nvidia shares have shed 7.0% over seven straight sessions, and the sell-off has little to do with waning appetite for AI chips. The harder question facing investors is whether the orders piling up can be financed, powered and switched on fast enough to justify the multiple.
Chief executive Jensen Huang insists demand still outruns supply. Speaking at an event in Berlin, he argued that the long-run returns on AI hardware investment remain intact and framed the current turbulence as temporary, according to Bloomberg. New orders, he added in the BG2 podcast published Thursday, arrive almost weekly and sometimes daily, with the largest customers planning as much as two years ahead. Cloud giants, AI cloud operators and device makers are all placing sizeable bookings.
What stands between those bookings and revenue is not willingness to buy. Huang pointed to electricity shortages, unfinished data centers and missing manufacturing components as the binding constraints on how quickly capacity can be brought online. That distinction matters: demand and deliverable revenue are separate variables, and delays in power supply or construction could damp near-term growth even against a fat order book.
Two Ways to Read the Slide
The seven-day decline admits two interpretations. It could represent a cheaper entry point into a company whose demand has not wavered — or it could be the market pricing in a slower conversion of orders into shipped hardware and booked sales. Huang's remarks support the first reading without ruling out the second.
The bull case does not require another surge in market euphoria. It requires only that Nvidia gradually work through the backlog, in which case today's capacity squeeze looks like a timing issue rather than a demand problem. DBS chief investment officer Hou Wey Fook expects 70% earnings growth from Nvidia next year and puts the stock at 17 times forward twelve-month earnings, arguing AI shares remain far from bubble territory. That combination only convinces if the projected growth actually materializes — a valuation built on future profits draws its meaning from their delivery.
Should investors sell immediately? Or is it worth buying Nvidia?
Breadth of buyers reinforces the argument. Demand spanning hyperscalers, AI cloud providers and equipment manufacturers suggests the build-out is not confined to a handful of projects, so different customer groups could carry expansion even if some initiatives start later than planned.
CoreWeave Shows Where the Skepticism Lives
The bear case is not that AI chips suddenly go unwanted. It is that elevated financing costs and insufficient project returns could throttle the build-out — a concern Reuters reports is circulating on Wall Street, with debt-funded infrastructure under particular strain. Saxo chief investment strategist Charu Chanana sees the AI sector splitting more sharply between winners and losers, with investors demanding profitability, viable financing and capital discipline.
CoreWeave illustrates the exposure. Reuters puts the company's debt at roughly $50 billion, with capital expenditure projected at $35–39 billion for 2026. Nvidia, a significant CoreWeave shareholder, has committed to purchase unsold capacity through April 2032 — a backstop that ties the chipmaker directly to how well that infrastructure is utilized. Shares and bonds of CoreWeave were headed for steep weekly losses, and the funding strain has spilled across the AI sector.
Higher-than-expected US inflation data for September, reported Wednesday, pushed Treasury yields up and added further pressure. Should tighter credit collide with doubts about AI's economic payoff, delayed customer projects and financial stress could arrive together — leaving large orders as evidence of long-term need without safeguarding near-term growth.
Beyond Raw Compute: Design Tools and Optical Links
Nvidia's ambitions extend past selling more accelerators. The company said Wednesday that semiconductor firms are using its CUDA-X libraries for AI-assisted chip design and manufacturing, including a new library called cuPotential, with applications spanning design, fabrication and inspection. The wider the adoption, the more Nvidia could participate in ongoing development and production workflows rather than a single step of the chain.
A separate collaboration announced Monday with Cisco targets an optical module format for next-generation AI factories, supporting 400 Gb/s per lane and capacity of up to 12.8 Tb/s. The two efforts operate on different layers — software for chip development, planned connectivity hardware for AI infrastructure — and could prove complementary if they move from announcement to deployment. Neither eliminates financing risk, and confirmed adoption is not the same as a later revenue effect.
November 20 Is the Next Real Test
For now, a growth-oriented reading holds as long as demand exceeds supply and customers can finish their facilities. Additional delivery capacity would then help convert bookings into business, and the key disclosure would be not the volume of new orders but commentary on delivery capability and the pace at which customer projects come online.
If financing for major projects seizes up, earnings assumptions come under real pressure — and a lower forward multiple offers thin protection, since falling expectations can erode the very valuation advantage it appears to confer. Nvidia's results, due November 20, provide the next concrete checkpoint, with guidance on demand, capacity expansion and potential delays carrying the most weight. Progress on execution would give Huang's confidence an operational footing. Fresh obstacles would force the valuation to be measured against growth that arrives more slowly.
Ad
Nvidia Stock: New Analysis - 17 October
Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
