By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panel is largely bearish on Bloom Energy (BE) due to execution risks, cash burn, and project timing volatility, despite its impressive backlog and Oracle win. They are more divided on Oneok (OKE), with some seeing it as a reliable infrastructure play and others questioning its ability to survive the capital intensity of lateral pipeline buildout and hedging gas exposure. The real risk is the regulatory lag in grid interconnection, which could stall both companies.
Risk: Execution frictions such as grid interconnection, permitting, and supply chain issues, as well as the regulatory lag in grid interconnection.
Opportunity: Oneok's potential to command premium take-or-pay contracts for 'behind-the-meter' infrastructure and capture value at lower execution risk, as noted by Gemini.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Power has become one of the biggest bottlenecks in the global artificial intelligence (AI) build-out. Goldman Sachs expects the demand for power in the U.S. data center market to rise from 31 gigawatts (GW) in 2025 to 66 GW by 2027, driven mainly by the rapid expansion of AI infrastructure.
Hence, companies that provide the power and energy infrastructure supporting the AI economy, such as Bloom Energy (NYSE: BE) and Oneok (NYSE: OKE), can also prove to be smart picks. Here's why.
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
Bloom Energy's solid oxide fuel cells provide on-site power, which is becoming increasingly valuable for data-center developers facing electricity shortages and grid connection delays.
Bloom Energy is positioning itself as a core AI power supplier, not just a backup power vendor. Oracle's (NYSE: ORCL) Project Jupiter, an upcoming multi-gigawatt AI factory in New Mexico, will use up to 2.45 GW of power from Bloom Energy servers. That replaces previously planned gas turbines and backup diesel generators.
Additionally, more than half of Bloom Energy's current data-center backlog comes from contracts with other hyperscalers, AI-focused cloud providers, and colocation operators that lease data-center capacity to customers at the end of the first quarter of fiscal 2026 (ending March 31, 2026). The company also exited fiscal 2025 with roughly $20 billion of total current backlog.
The financial results are beginning to reflect that demand momentum. The company's revenue surged 130.4% year over year to $751.1 million. Management now expects full-year fiscal 2026 revenue to fall in the range of $3.4 billion to $3.8 billion.
However, Bloom Energy is also exposed to significant project-timing risk. Shares recently fell nearly 10%, after reports that construction was paused at a 1.8 GW Crusoe Energy data center project, which also involved Bloom Energy. Hence, the company now needs to demonstrate that its large backlog can be converted into revenue without major delays.
Oneok is a major midstream energy company that transports, processes, stores, and exports natural gas liquids, natural gas, refined products, and crude oil.
Oneok's infrastructure is becoming more relevant, as data centers increase demand for natural gas-fired power. The company is in advanced discussions with data center customers in Oklahoma and Texas. The company is also evaluating more than 40 data-center-related counterparties representing more than 5 billion cubic feet per day of potential natural gas demand. Some data-center projects that were initially expected to be small pipeline connections have grown into larger opportunities, as hyperscalers now require bigger gas volumes and larger pipelines.
Four leading AI models discuss this article
"Bloom Energy's project-timing risk outweighs its AI backlog narrative until at least two large deployments reach commercial operation without further pauses."
The article correctly flags surging AI-driven power demand but underplays Bloom Energy's acute execution risks. Its $20B backlog and Oracle deal sound impressive, yet a single 1.8 GW project pause already triggered a 10% drop, and converting multi-GW fuel-cell orders into revenue by FY2026 faces grid, permitting, and supply-chain hurdles that could stretch timelines. Oneok's gas-volume discussions look more durable given its midstream footprint, yet even there hyperscaler demand remains speculative and competes with renewables plus efficiency gains. Fiscal 2025 revenue doubling to $751M is real, but margins and cash conversion on large projects are unproven at this scale.
Bloom's solid-oxide tech could win on-site power mandates faster than expected if grid delays worsen, turning the current backlog into 2026-27 revenue beats that justify a re-rating.
"The AI data-center power story hinges on aggressive backlog-to-revenue conversion and favorable energy-cost dynamics; delays or cost shocks could derail the expected multi-year returns."
The piece pitches Bloom Energy (BE) and Oneok (OKE) as AI-era power backbones, anchored by Goldman’s view that data-center power demand could rise from 31 GW in 2025 to 66 GW by 2027. Bloom shows a roughly $20B backlog with revenue up 130% YoY to $751M and guidance of $3.4–3.8B for FY2026; Oneok stands to profit from growing gas-fired data-center demand and discussions with 40+ counterparties. Yet the thesis hinges on execution and timing: Crusoe Energy’s 1.8 GW pause highlights project-delivery risk, while BE’s gas/hydrogen cost exposure and OKE’s regulatory/commodity volatility could cap returns despite a big backlog.
Bullish counterpoint: if the AI capex cycle remains durable, hyperscalers will push large-scale deployments forward and backlog conversion could surprise to the upside. The Crusoe pause might be a timing hiccup rather than a fundamental risk, and BE/OKE could still capture meaningful margin expansion as scale grows.
"The AI power trade is bifurcated: OKE offers defensive cash-flow stability, while BE remains a high-risk, high-reward bet on the failure of traditional grid infrastructure."
The article correctly identifies the power bottleneck, but conflates two distinct risk profiles. Bloom Energy (BE) is a speculative play on distributed generation; while the Oracle win is a massive proof-of-concept, their history of cash burn and project-timing volatility makes them a 'show-me' story rather than a reliable infrastructure play. Conversely, Oneok (OKE) is a classic midstream utility play; their integration into the AI trade is a margin-expansion story, not a growth story. Investors should focus on OKE's ability to monetize existing pipelines without heavy capex, while treating BE as a high-beta option on grid failure. The real risk is the regulatory lag in grid interconnection, which could stall both.
The thesis assumes AI data centers will exclusively rely on private, on-site power, ignoring the likelihood that hyperscalers will prioritize direct, low-cost utility-scale nuclear or renewable PPA deals that could render Bloom's fuel cells economically uncompetitive.
"AI power demand is real, but BE faces acute project-timing risk and OKE's midstream margins are structurally compressed by hyperscaler negotiating power—neither justifies a 'best stocks to buy' label without clarity on unit economics and contract terms."
The article conflates two distinct theses: (1) AI power demand is real and growing—Goldman's 31→66 GW forecast is plausible; (2) BE and OKE are optimal vehicles to capture it. On (1), I'm convinced. On (2), I'm skeptical. BE trades on backlog conversion risk the article acknowledges but minimizes—a 1.8 GW pause is material when your total backlog is ~$20B. OKE's 40+ counterparties and 5 BCF/day potential sounds impressive until you ask: how many actually build, and at what margin? Midstream typically earns 5–8% EBITDA margins on throughput; hyperscalers will demand long-term fixed-price contracts, compressing upside. The article also ignores that utilities (NEE, DUK) and pure-play power REITs (CWEN, AEP) may capture more durable value at lower execution risk.
If BE's backlog converts even 60% on schedule and OKE locks in 2+ BCF/day at acceptable rates, both could re-rate 40%+ over 18 months—and the article's omission of execution risk may simply reflect that Street consensus already prices it in.
"Grid delays create a temporary on-site advantage for BE that utilities cannot match until reforms accelerate."
Claude flags utilities like NEE and DUK as lower-risk captures, but this underweights the 3-5 year interconnection queues that Gemini separately notes. Those delays hand BE a genuine 2025-27 window for on-site deployments even at premium pricing, provided the Oracle proof-of-concept scales without the cash-burn repeats that have dogged prior cycles.
"Execution friction and grid/permit delays will cap BE/OKE's near-term upside; backlog alone won't drive margins without faster approvals and on-site deployment."
Claude’s skepticism about BE/OKE is reasonable, but the real bottleneck is not just backlog size, it’s execution frictions—grid interconnection, permitting, supply chain—that 1.8 GW Crusoe pause signals. Even with Oracle’s proof, revenue won’t flow until regulators green-light and contractors mobilize. That timing risk cements near-term cash burn and muddies margins, meaning a re-rating relies on a material acceleration in approvals and on-site deployments, not just backlog management.
"Oneok's value lies in its ability to secure premium, long-term take-or-pay contracts with hyperscalers that transcend traditional midstream margin caps."
Claude, your focus on midstream EBITDA margins misses the structural shift: hyperscalers are essentially becoming utility-scale off-takers, not just throughput customers. Oneok isn't just selling gas; they are positioning for 'behind-the-meter' infrastructure where they can command premium take-or-pay contracts that bypass traditional utility regulatory lag. The risk isn't margin compression; it's the capital intensity required to build out the lateral pipeline capacity to data center sites before the grid catches up.
"Take-or-pay contracts shift capex risk to OKE without necessarily improving returns if hyperscalers demand long-term fixed rates."
Gemini's 'behind-the-meter' take on OKE is clever, but take-or-pay contracts don't solve the capex problem—they just shift risk to OKE's balance sheet. If hyperscalers demand 10-year fixed rates to lock in data-center economics, OKE absorbs commodity and demand volatility. That's not premium pricing; it's leverage against OKE. The real question: does OKE's balance sheet survive 2-3 GW of lateral buildout at 5-8% returns while hedging 10-year gas exposure?
The panel is largely bearish on Bloom Energy (BE) due to execution risks, cash burn, and project timing volatility, despite its impressive backlog and Oracle win. They are more divided on Oneok (OKE), with some seeing it as a reliable infrastructure play and others questioning its ability to survive the capital intensity of lateral pipeline buildout and hedging gas exposure. The real risk is the regulatory lag in grid interconnection, which could stall both companies.
Oneok's potential to command premium take-or-pay contracts for 'behind-the-meter' infrastructure and capture value at lower execution risk, as noted by Gemini.
Execution frictions such as grid interconnection, permitting, and supply chain issues, as well as the regulatory lag in grid interconnection.