Nebius Adds Inference Technology, but Valuation Leaves Little Margin

Adobe's fiscal third quarter offered evidence that customers are using its AI products at scale. It also left investors with a harder question: how quickly will that usage convert into durable contracted revenue?

Revenue rose 13% as reported to $6.76 billion, and non-GAAP earnings per share reached $6.13. Adobe raised its full-year targets to $26.576 billion to $26.626 billion of revenue and $24.45 to $24.50 of non-GAAP EPS.

The strongest operating signal came from AI adoption. AI-first annual recurring revenue exceeded $650 million and grew more than 150% from a year earlier. Firefly ending ARR from the app and credit packs increased 40% from the prior quarter. Creative freemium monthly active users passed 100 million, up more than 70%.

Those figures show reach, but they measure different stages of the funnel. Freemium users indicate product exposure. Credit consumption shows activity. ARR shows contracted or recurring monetization. Investors need all three to rise without heavy discounting or damage to Adobe's high-margin subscription model.

The counterpoint is remaining performance obligations, which grew 8% to $22.16 billion. Current RPO rose 9%. Both were slower than reported revenue growth, which can indicate that the future contracted base is not accelerating as quickly as current usage. It is not a forecast of decline, but it explains why the market remains skeptical despite solid quarterly results.

Cash generation stayed strong. Operating cash flow reached a quarterly record of $2.52 billion, and Adobe repurchased about 9.5 million shares. The company had approximately $24.55 billion left under its April 2026 buyback authorization.

Leadership is also changing. Anil Chakravarthy becomes president and chief executive on Dec. 1, while Shantanu Narayen moves to executive chair. Chakravarthy inherits a company with a large installed base and fast-growing AI engagement, but also pressure to prove that generative tools expand Adobe's economics rather than commoditize creative work.

The next phase is therefore about conversion. Adobe must turn free users into paying customers, grow Firefly and agentic products without eroding core pricing, and lift net new ARR. The quarter showed that AI demand exists inside Adobe's ecosystem. It did not yet settle how valuable that demand will be.

Gartner expects more than 80% of neocloud providers to cease operating independently by 2030. The forecast does not say that AI computing demand will disappear. It says many specialized providers may not earn enough, quickly enough, to finance the assets needed to serve that demand.

The research firm estimates AI providers will spend nearly $1 trillion on infrastructure in 2026 and another $6 trillion from 2027 through 2030. Gartner calculates that data-center owners would need $8.5 trillion of cumulative revenue to cover announced capital spending, financing costs and a 12% return on invested capital during that period. It forecasts no more than $3.6 trillion of attributable enterprise and consumer AI revenue.

Neoclouds are especially exposed because their revenue base is narrower than that of Amazon, Microsoft or Alphabet. A hyperscaler can fund data centers with advertising, software, commerce or other cloud cash flow. A specialized GPU cloud often depends on debt, customer prepayments, equity issuance and continued access to capital markets.

CoreWeave illustrates both sides. Its second-quarter revenue more than doubled to $2.58 billion and backlog reached about $104 billion, showing strong contracted demand. Net interest expense also rose to $640 million, and the company continued to raise large amounts of debt and convertible capital. The question is not whether customers want compute. It is whether cash arrives before financing costs and hardware obsolescence consume the return.

Gartner expects spending on AI-optimized infrastructure services to rise from $66 billion in 2027 to $174 billion in 2030. That is substantial growth, but the firm argues it may still be too slow for many providers to recover investment before the current generation of accelerators is replaced.

Ceasing independent operations can mean several outcomes: a sale, merger, restructuring or shutdown. The servers and data centers do not necessarily disappear. Better-capitalized buyers may acquire them at lower prices, while lenders and shareholders absorb part of the loss.

That makes balance-sheet structure a competitive advantage. Investors should compare contracted revenue with debt maturities, interest expense, customer concentration, useful-life assumptions and the amount of capital still required. High demand can coexist with poor returns when the asset base is expensive and depreciates quickly.

The Gartner forecast is a scenario, not a count of companies destined to fail. Its value is the framework: in AI infrastructure, the scarce resource may be affordable long-duration capital rather than customer interest.

Rocket Lab's proposed $8 billion acquisition of Iridium has drawn a competition challenge at the Federal Communications Commission, adding another uncertainty to a company already managing a delay to its Neutron rocket.

Ligado Networks asked the FCC to examine whether Iridium has used regulatory filings to slow competing satellite services. Ligado has a direct commercial interest: its planned SkyTerra Next service with AST SpaceMobile would use satellite connectivity for direct-to-device communications and could compete with Iridium.

The filing does not amount to an FCC finding. It is an allegation from an interested party in an active proceeding, and Rocket Lab and Iridium will have an opportunity to respond. Investors should distinguish that process from a rejection or a formal enforcement action.

The underlying transaction is much larger than a spectrum dispute. Rocket Lab agreed in June to pay $54 per Iridium share in cash and stock, valuing the company at about $8 billion including debt. Iridium brings a global L-band network, more than 2.55 million active subscribers, over 500 distribution partners and recurring communications revenue. Rocket Lab would add those assets to launch, spacecraft manufacturing and satellite components.

That vertical integration could lower launch costs for a future Iridium constellation and give Rocket Lab immediate exposure to direct-to-device, Internet of Things and positioning services. It also increases execution risk. The deal requires regulatory approvals, financing and the integration of a profitable network with a capital-intensive launch company.

Neutron makes the timing harder. Rocket Lab moved the medium-lift rocket's first flight into 2027 after identifying manufacturing defects in a carbon-composite propellant-tank component. Neutron is central to the logic of launching larger constellations internally, so further delays could postpone some expected benefits even if the acquisition closes.

Rocket Lab shares had fallen about 25% over the prior week and 38% from the Oct. 7 close reported by the source. Price action reflects both issues, but they should be monitored separately. Neutron is an engineering and schedule risk. The Iridium proceeding is a regulatory and competitive risk.

The FCC challenge is not proof that the transaction will fail. It does mean the public-interest case will receive closer scrutiny, especially around spectrum and competitor access. For investors, the next evidence is the applicants' response, the FCC timetable and whether Neutron's revised tests support a credible 2027 launch plan.

SpaceX's agreement to buy nationwide 800 MHz spectrum from Grain Management has forced investors to revisit a threat that once looked distant: satellite service becoming a meaningful alternative to terrestrial wireless.

The reported transaction value was about $8 billion, and the purchase still requires Federal Communications Commission approval. The spectrum alone does not create a national mobile network. SpaceX would also need satellites, ground infrastructure, devices and substantial capital before Starlink Mobile could compete broadly with Verizon, AT&T or T-Mobile.

The strategic direction is still clear. Lower-band spectrum travels farther and penetrates buildings better than higher-frequency signals, making it valuable for wide-area coverage. SpaceX has said upgraded Starlink Mobile service using next-generation satellites is targeted for the end of 2027. Even a limited service could affect carrier pricing if it becomes credible in rural coverage, emergency connectivity or wholesale arrangements.

Verizon enters that contest with a mixed financial profile. Its 2025 operating revenue was $138.2 billion, up from $134.8 billion in 2024, and free cash flow was $20.1 billion. Total unsecured debt was $131.1 billion at year-end, although net unsecured debt declined to $110.1 billion and the company's net unsecured debt to adjusted EBITDA ratio was 2.2 times. Those figures are more precise than treating every liability as operating debt.

The Frontier acquisition is Verizon's most important answer. The deal closed on Jan. 20 and expanded the combined network to roughly 30 million fiber passings. Fiber lets Verizon bundle home broadband with mobile service, potentially reducing churn and giving households a reason to keep multiple services with one provider.

That defense must show up in operating results. Verizon guided to 2% to 3% growth in mobility and broadband service revenue for 2026, at least $21.5 billion of free cash flow and a fiber build pace of at least 2 million passings. Meeting those goals would give the company room to invest while supporting its dividend.

SpaceX is not likely to take a large share of wireless subscribers quickly. The nearer risk is that its presence changes negotiations before it changes market share, pressuring wholesale economics or forcing carriers to spend more. For Verizon, the differentiator is no longer a simple claim of having the best mobile network. It is whether the company can combine wireless, fiber and service quality into a package that satellite connectivity cannot easily replace.

ARK Invest bought 132,498 CoreWeave shares on Oct. 19, adding to the AI cloud provider after a decline that had pushed the stock toward a third straight losing week. The purchase split between 102,280 shares for ARKK and 30,218 for ARKW, worth about $8.1 million at the reported trade prices.

The trade is notable, but it does not resolve the question weighing on CoreWeave: how much of its growth will accrue to shareholders after financing the infrastructure behind it.

CoreWeave reported $2.58 billion of revenue for the second quarter of 2026, more than double the year-earlier figure, and a revenue backlog of about $104 billion. Demand is not the immediate problem. The pressure sits below the operating line. Net interest expense rose to $640 million from $267 million a year earlier, and the company posted a $626 million net loss.

CoreWeave is financing data centers and accelerators years before much of the contracted revenue arrives. That structure can work when capacity remains scarce, customers honor long commitments and refinancing markets stay open. It becomes less forgiving when interest rates rise or investors demand more protection.

The company has tried to broaden its funding base. In March, it closed an $8.5 billion delayed-draw facility backed by GPU infrastructure and a customer contract, which received investment-grade ratings from Moody's and DBRS. In September, it proposed $3 billion of convertible notes. Those transactions show that CoreWeave still has market access, but they also underline how central outside capital is to the expansion plan.

Analysts are divided for the same reason. Goldman Sachs upgraded the shares to Buy, arguing that tight AI capacity can support attractive returns. MoffettNathanson and BNP Paribas cut price targets, focusing on obligations and the risk that financing becomes more expensive.

ARK's purchase is therefore a vote on more than AI demand. It is a bet that CoreWeave can convert a large backlog into cash before its capital structure becomes the limiting factor. The next useful evidence will be cash generation, interest coverage and the terms of new financing, not the size of one fund manager's trade.