Who pays for AI’s power bill? A new pledge — or political theater?

Last week’s State of the Union brought the surprising image of the president leaning into the very modern problem of AI data centers and electricity rates. He announced a “rate payer protection pledge” and said major tech companies would sign deals next week to “provide for their own power needs” so local electricity bills don’t spike. It sounds neat: hyperscalers build or buy their own power, communities don’t pay more, and everybody moves on. But the reality is messier — and more revealing about how energy, politics, and tech interact.

What was announced — in plain English

• President Trump announced during the February 24, 2026 State of the Union that the administration negotiated a “rate payer protection pledge.” (theverge.com)
• The White House said major firms — Amazon, Google, Meta, Microsoft, xAI, Oracle, OpenAI and others — would formally sign a pledge at a March 4 meeting to shield ratepayers from electricity price increases tied to AI data-center growth. (foxnews.com)
• The administration framed the fix as letting tech companies build or secure their own generation (including new power plants) so the stressed grid doesn’t force higher bills on surrounding communities. (theverge.com)

Why this matters now

• AI data-center construction and operations have grown fast, pulling large blocks of power and creating hot local debates about grid strain, rates, and environmental impacts. Utilities and state regulators often negotiate special rates or infrastructure upgrades for big customers — which can shift costs around. (techcrunch.com)
• Politically, energy costs are a live issue for voters. A presidential pledge that promises to blunt rate increases is attractive even if the mechanics are complicated. Axios and Reuters noted the move’s symbolic weight. (axios.com)

How much of this is new versus PR?

• Much of the headline pledge echoes commitments big cloud providers have already made: signing deals to buy or build generation, increasing efficiency, and in some cases directly investing in local energy projects. Companies such as Microsoft have already offered community-first infrastructure plans in some locations. So the White House announcement amplifies existing industry steps rather than inventing a wholly new approach. (techcrunch.com)
• Legal and logistical constraints matter. Electricity markets and permitting sit mostly at state and regional levels, and the federal government can’t unilaterally force a nationwide energy-market restructuring. A White House-hosted pledge can add political pressure, but enforcement and the details of cost allocation remain in many hands beyond the president’s. (axios.com)

Practical questions that matter (and aren’t answered yet)

• Who pays up front? If a company builds generation, does it absorb the capital cost entirely, or does it receive tax breaks, subsidies, or other incentives that effectively shift some burden back to taxpayers? (nextgov.com)
• What counts as “not raising rates”? If a company signs a pledge to “not contribute” to local bill increases, regulators will still need to verify causation and fairness across customer classes.
• Will companies build fossil plants, gas peakers, renewables, or pursue grid-scale battery and demand-response strategies? The administration has signaled support for faster fossil-fuel permitting, which would shape outcomes. (theverge.com)

The investor and community dilemma

• For local officials and residents, a tech company saying “we’ll pay” is appealing — but communities still face issues of water use, land use, emissions, and long-term tax and workforce impacts that a power pledge doesn’t fully resolve. (energynews.oedigital.com)
• For energy markets and utilities, the ideal outcome is coordinated planning: companies that participate in grid upgrades, pay cost-reflective rates, and contract for incremental generation or storage reduce scramble-driven rate spikes. That coordination is harder than a headline pledge. (techcrunch.com)

What to watch next

• The March 4 White House meeting: who signs, and what are the actual commitments (capital investments, long-term purchase agreements, operational guarantees, or merely statements of intent). (cybernews.com)
• State regulatory responses: states with recent data-center booms (and local rate concerns) may adopt rules or require formal binding commitments from developers. (axios.com)
• The type of generation and permitting choices: promises to “build power plants” can mean very different environmental and fiscal outcomes depending on whether those plants are gas, renewables, or nuclear. (theverge.com)

Quick wins and pitfalls

• Quick wins: companies directly investing in local grid upgrades, long-term power purchase agreements (PPAs) tied to new renewables plus storage, and transparent cost-sharing with local utilities can reduce friction. (techcrunch.com)
• Pitfalls: vague pledges without enforceable terms; incentives that mask public subsidies; and a federal play that ignores regional market rules could leave communities still paying the tab indirectly. (axios.com)

My take

This announcement will matter most if it turns political theater into enforceable, transparent commitments that prioritize community resilience and low-carbon options. Tech companies already have incentives — reputation, permitting ease, and long-term operational stability — to address their power footprint. The White House pledge can accelerate those moves, but it shouldn’t be a substitute for thorough state-level regulation, utility planning, and honest accounting of who pays and who benefits.

If the March 4 signings produce detailed, binding contracts (with measurable timelines, public reporting, and third-party oversight), this could be a meaningful pivot toward smarter energy planning around AI. If they’re broad press statements, expect headlines — and continuing fights at city halls and public utility commissions.

Sources

• Trump claims tech companies will sign deals next week to pay for their own power supply. The Verge. https://www.theverge.com/science/884191/ai-data-center-energy-state-of-the-union-trump. (theverge.com)
• TechCrunch: The White House wants AI companies to cover rate hikes. Most have already said they would. https://techcrunch.com/2026/02/25/the-white-house-wants-ai-companies-to-cover-rate-hikes-most-have-already-said-they-would/. (techcrunch.com)
• Reuters / Energy News: White House hosts Big Tech to pledge to reduce power costs. https://energynews.oedigital.com/environment/2026/02/25/white-house-hosts-big-tech-to-pledge-to-reduce-power-costs. (energynews.oedigital.com)
• Axios: Tech ratepayer pledge may be more bark than bite – but still matters. https://www.axios.com/2026/02/26/big-tech-ratepayer-pledge-trump. (axios.com)
• CBS News State of the Union highlights (excerpt on data-center energy deal). https://www.cbsnews.com/live-updates/state-of-the-union-2026/. (cbsnews.com)

Moonshots and Mutinies: Elon Musk Wants a Lunar Factory to Launch AI Satellites

The headline sounds like science fiction: build a factory on the Moon, assemble AI satellites there, then fling them into orbit with a giant catapult. But this is exactly the vision Elon Musk sketched for xAI at a recent all‑hands meeting — a talk first reported by The New York Times and covered by TechCrunch and other outlets. The timing is notable: co‑founders departing, a major reorg, and a SpaceX‑xAI merger that some expect will lead to a blockbuster IPO later this year. The result is a mix of bravado, engineering fantasy, strategic logic, and regulatory questions — the kind of story that forces you to ask whether this is grand strategy or grandstanding.

Why this matters now

• xAI is freshly merged into Elon Musk’s space and social empire, amplifying ambitions and tightening the spotlight.
• Several of xAI’s original co‑founders have recently left, raising questions about execution and culture during a pivotal scaling phase.
• Musk’s moon plan reframes the debate about where the future of compute will live — on Earth, in orbit, or on the lunar surface — and what would be required to get there.

The pitch in plain language

According to reporting summarized by TechCrunch, Musk told xAI employees that:

• xAI will need a lunar manufacturing facility to build AI satellites.
• The proposed lunar facility would include a mass driver — an electromagnetic catapult — to launch satellites into space.
• The rationale is raw compute scale: the Moon (and space in general) offers a way to access vast energy and cooling potential that Earth datacenters can’t match.

Those comments came during an all‑hands that coincided with a flurry of departures by co‑founders such as Tony Wu and Jimmy Ba, and as the merged entity prepares for a possible IPO. TechCrunch later published the full 45‑minute all‑hands video, which adds context to the public reporting.

Why a lunar factory sounds plausible (on paper)

• Energy and cooling: Space (and the lunar surface) offers unique opportunities, e.g., direct access to sunlight for massive solar farms and passive cooling in shaded regions — appealing for power‑hungry AI clusters.
• Vertical integration: Musk’s conglomerate already spans rockets (SpaceX), social/data platforms (X), and energy/transport (Tesla, Starlink synergies). Adding lunar manufacturing could be pitched as the next step in controlling a full stack of data, transport, and infrastructure.
• Proprietary data and differentiation: A moon‑based platform could, in theory, enable data flows and sensors unavailable to competitors — feeding a unique “world model” that Musk has described as the long‑term objective.

The big, practical hurdles

• Engineering scale: Building habitable factories, reliable lunar construction techniques, and a functional mass driver are orders of magnitude harder than launching satellites from Earth. Cost, time, and risk are enormous.
• Legal and geopolitical limits: The 1967 Outer Space Treaty bars national appropriation of celestial bodies. U.S. law allows companies to extract resources they mine, but the legal landscape for permanent facilities and mass industrial activity is contested internationally.
• Talent and timing: Key technical leaders exiting during a reorg makes execution riskier. Ambitious long‑horizon projects don’t mesh easily with the short timelines and accountability of public markets and IPO cycles.
• Environmental and safety concerns: Unproven large‑scale lunar manufacturing and mass drivers raise questions about space debris, lunar environment stewardship, and collision risk for satellites and crewed missions.

What investors and competitors see

• Investors may cheer the vision’s upside: unique assets and defensible moats that could justify sky‑high valuations if achieved.
• Shorter time‑horizon stakeholders (public markets, customers, partners) will want tangible milestones: product roadmaps, revenue paths, and credible technical milestones long before any lunar steel is laid.
• Competitors are watching the tech stack: if the Moon pitch is an attempt to lock in energy, data, and unique sensors, rivals will adapt via orbital compute, international partnerships, or legal/policy pressure.

A few scenarios to watch

• Near term (months): continued reorg and talent churn at xAI; more public messaging to frame the Moon idea as long‑term strategy rather than an immediate product pivot.
• Medium term (1–3 years): concrete engineering programs announced — prototypes for orbital data centers, power projects, or lunar robotics partnerships — which would signal movement from concept to execution.
• Long term (decades): if the idea survives technical, legal, and funding hurdles, it could reshape where large AI clusters live — and who controls the data those clusters consume.

Notes on credibility and context

• TechCrunch’s coverage and the publicly posted all‑hands video are non‑paywalled, accessible records of the pitch and surrounding company changes.
• Reporting across outlets (The Verge, Financial Times, TechCrunch) shows consistent core claims: Musk pitched lunar infrastructure as part of xAI’s future while several co‑founders departed.
• Some outlets add detail or editorial framing (e.g., energy scale ambitions, concerns about deepfakes on X), which are relevant to the company’s near term optics but separate from the moon manufacturing claim itself.

What this says about Musk’s strategy

• Moon plans are less a literal product roadmap than a narrative lever: they signal scale, ambition, and an integrated multi‑domain approach that stokes investor enthusiasm.
• The vision ties disparate pieces of Musk’s empire into a single storyline: rockets, satellites, social data, and energy converge into a proprietary vertical. That’s strategically coherent — if technically audacious.
• For employees and early leaders, the shift from a scrappy startup to a multi‑domain industrial ambition means differing skill sets and appetites for risk — which helps explain departures amid reorganization.

My take

There’s a productive tension here between audacity and accountability. Big visions — even wildly improbable ones — have a role in attracting capital and talent. But the moment you promise lunar factories and mass drivers, you invite intense scrutiny: technical feasibility, timelines, legal permission, and human capital. The most useful question for xAI and its stakeholders is not whether the Moon is “possible” in a vacuum; it’s whether the company can credibly deliver meaningful intermediate milestones that justify investment and retain top talent while the moonshot remains decades away.

Final thoughts

Ambition keeps technology moving forward, but execution makes it real. Musk’s lunar pitch is headline‑grabbing and strategically provocative; whether it becomes a blueprint or a branding exercise depends on the hard, incremental work that follows: prototypes, partnerships, regulatory clarity, and, crucially, people who stay to build it.

Sources

• With co‑founders leaving and an IPO looming, Elon Musk turns talk to the moon | TechCrunch
  https://techcrunch.com/2026/02/10/with-co-founders-leaving-and-an-ipo-looming-elon-musk-turns-talk-to-the-moon/

• xAI lays out interplanetary ambitions in public all‑hands | TechCrunch
  https://techcrunch.com/2026/02/11/xai-lays-out-interplanetary-ambitions-in-public-all-hands/

• Two more xAI co‑founders are among those leaving after the SpaceX merger | The Verge
  https://www.theverge.com/ai-artificial-intelligence/877609/two-more-xai-co-founders-are-among-those-leaving-after-the-spacex-merger

When one AI cites another: ChatGPT, Grokipedia and the risk of AI-sourced echo chambers

Information wants to be useful — but when the pipes that deliver it start to loop back into themselves, usefulness becomes uncertain. Last week’s revelation that ChatGPT has begun pulling answers from Grokipedia — the AI-generated encyclopedia launched by Elon Musk’s xAI — isn’t just a quirky footnote in the AI wars. It’s a reminder that where models get their facts matters, and that the next chapter of misinformation might not come from trolls alone but from automated knowledge factories feeding each other.

Why this matters right now

• Grokipedia launched in late 2025 as an AI-first rival to Wikipedia, promising “maximum truth” and editing driven by xAI’s Grok models rather than human volunteer editors.
• Reporters from The Guardian tested OpenAI’s GPT-5.2 and found it cited Grokipedia multiple times for obscure or niche queries, rather than for well-scrutinized topics. TechCrunch picked up the story and amplified concerns about politicized or problematic content leaking into mainstream AI answers.
• Grokipedia has already been criticized for controversial content and lack of transparent human curation. If major LLMs start using it as a source, users could get answers that carry embedded bias or inaccuracies — with the AI presenting them as neutral facts.

What happened — a short narrative

• xAI released Grokipedia in October 2025 to great fanfare and immediate controversy; some entries and editorial choices were flagged by journalists as ideological or inaccurate.
• The Guardian published tests showing that GPT-5.2 referenced Grokipedia in several responses, notably on less-covered topics where Grokipedia’s claims differed from established sources.
• OpenAI told reporters it draws from “a broad range of publicly available sources and viewpoints,” but the finding raised alarm among researchers who worry about an “AI feeding AI” dynamic: models trained or evaluated on outputs that themselves derive from other models.

The risk: AI-to-AI feedback loops

• Repetition amplifies credibility. When a large language model cites a source — and users see that citation or accept the answer — the content’s perceived authority grows. If that content originated from another model rather than vetted human scholarship, the process can harden mistakes into accepted “facts.”
• LLM grooming and seeding. Bad actors (or even well-meaning but sloppy systems) can seed AI-generated pages with false or biased claims; if those pages are scraped into training or retrieval corpora, multiple models can repeat the same errors, creating a self-reinforcing echo.
• Loss of provenance and nuance. Aggregating sources without clear provenance or editorial layers makes it hard to know whether a claim is contested, subtle, or discredited — especially on obscure topics where there aren’t many independent checks.

Where responsibility sits

• Model builders. Companies that train and deploy LLMs must strengthen source vetting and transparency, especially for retrieval-augmented systems. That includes weighting human-curated, primary, and well-audited sources more heavily.
• Source operators. Sites like Grokipedia (AI-first encyclopedias) need clearer editorial policies, provenance metadata, and visible mechanisms for human fact-checking and correction if they want to be treated as reliable references.
• Researchers and journalists. Ongoing audits, red-teaming and independent testing (like The Guardian’s probes) are essential to surface where models are leaning on questionable sources.
• Regulators and platforms. As AI content becomes a larger fraction of web content, platform rules and regulatory scrutiny will increasingly shape what counts as an acceptable source for widespread systems.

What users should do today

• Ask for sources and check them. When an LLM gives a surprising or consequential claim, look for corroboration from reputable human-edited outlets, primary documents, or scholarly work.
• Be extra skeptical on obscure topics. The reporting found Grokipedia influencing answers on less-covered matters — exactly the places where mistakes hide.
• Prefer models and services that publish retrieval provenance or let you inspect the cited material. Transparency helps users evaluate confidence.

A few balanced considerations

• Not all AI-derived content is inherently bad. Automated systems can surface helpful summaries and surface-level context quickly. The problem isn’t automation per se but opacity and lack of corrective human governance.
• Diversity of sources matters. OpenAI’s claim that it draws on a range of publicly available viewpoints is sensible in principle, but diversity doesn’t replace vetting. A wide pool of low-quality AI outputs is still a poor knowledge base.
• This is a systems problem, not a single-company scandal. Multiple major models show signs of drawing from problematic corners of the web — the difference will be which organizations invest in safeguards and which don’t.

Things to watch next

• Will OpenAI and other major model providers adjust retrieval weightings or add filters to downrank AI-only encyclopedias like Grokipedia?
• Will Grokipedia publish clearer editorial processes, provenance metadata, and human-curation layers to be treated as a responsible source?
• Will independent audits become standard industry practice, with third-party certifications for “trusted source” pipelines used by LLMs?

My take

We’re watching a transitional moment: the web is shifting from pages written by people to pages largely created or reworded by machines. That shift can be useful — faster updates, broader coverage — but it also challenges the centuries-old idea that reputable knowledge is rooted in accountable authorship and transparent sourcing. If we don’t insist on provenance, correction pathways, and human oversight, we risk normalizing an ecosystem where errors and ideological slants are amplified by the very tools meant to help us navigate information.

In short: the presence of Grokipedia in ChatGPT’s answers is a red flag about data pipelines and source hygiene. It doesn’t mean every AI answer is now untrustworthy, but it does mean users, builders and regulators need to treat the provenance of AI knowledge as a first-class problem.

Sources

• Latest ChatGPT model uses Elon Musk’s Grokipedia as source, tests reveal — The Guardian
  https://www.theguardian.com/technology/2026/jan/24/latest-chatgpt-model-uses-elon-musks-grokipedia-as-source-tests-reveal

• ChatGPT is pulling answers from Elon Musk’s Grokipedia — TechCrunch
  https://techcrunch.com/2026/01/25/chatgpt-is-pulling-answers-from-elon-musks-grokipedia/

One crushed the other: my take on ChatGPT‑5.1 vs Grok 4.1

The headline pretty much says it: after Tom’s Guide ran nine side‑by‑side prompts, one model didn’t just win — it dominated. If you’ve been following the weekly AI cage matches, this one matters because it shows where conversational AI is leaning: toward personality, interpretive depth, and emotional nuance.

Why this comparison matters

• Both ChatGPT‑5.1 and Grok 4.1 are among the most-talked‑about chatbots today.
• These are not incremental updates — they represent competing design philosophies: OpenAI’s emphasis on clarity, safety, and utility versus Grok’s (xAI/X) emphasis on boldness, candid tone, and contextual flair.
• A nine‑prompt shootout lets us see strengths and tradeoffs across categories that people actually care about: reasoning, creativity, humor, emotional support, and real‑world planning.

What the test looked at

Tom’s Guide used nine prompts spanning:

• Logic and trick questions
• Metaphors and explanations for kids
• Creative writing and storytelling
• Code generation and technical clarity
• Real‑world planning (travel iteneraries)
• Emotional intelligence and supportive messaging

The prompts were chosen to surface not just correctness but voice, subtext, and usefulness in everyday scenarios.

The short verdict

• Winner: Grok 4.1.
• Why: Grok took seven of the nine rounds, excelling at subtext, emotional tone, humor, and evocative creative writing. It was willing to call out trick questions, use more conversational slang when appropriate, and deliver answers that felt more human and expressive.
• ChatGPT‑5.1 wasn’t bad — it tended to be cleaner, more concise, and better at tightly constrained tasks (e.g., some concise metaphors and clean code), but it often felt more reserved compared with Grok’s bolder personality.

Highlights from the head‑to‑head

• Reasoning and trick questions
  • Grok flagged the classic “all but 9” puzzle as a trick and contextualized it; that extra metacognitive move won points for interpretive understanding.
• Creative writing and atmosphere
  • Grok built more tension and sensory detail in short fiction prompts; ChatGPT‑5.1 favored tighter structure and punchlines.
• Emotional support and tone
  • Grok used colloquial, authentic phrasing that resonated like a friend’s message — not “toxic‑positivity” but genuine validation. ChatGPT’s responses were supportive but more formal.
• Practical planning
  • ChatGPT‑5.1 sometimes won when the brief demanded balance, brevity, and modular practicality (e.g., family travel planning where flexibility matters).

What this tells us about AI design choices

• Personality vs. polish: Grok’s strength is personality. When human connection, subtext, or theatrical flair matters, personality wins. ChatGPT’s strength is polish: clarity, brevity, and predictability.
• Use‑case matters: If you want an assistant that’s a precise tool for structured tasks, the steadier, cleaner responses will be preferable. If your use case benefits from creative risk, humor, or raw empathy, a bolder voice can be more effective.
• The “best” model is context dependent: For developers, businesses, or educators, the ideal choice may combine the two approaches — or prefer one depending on brand voice and safety requirements.

Practical takeaways for users and creators

• Pick by outcome, not brand:
  • Need crisp instructions, safe defaults, or conservative language? Lean toward the model that favors clarity.
  • Want story mood, candid emotional replies, or punchy humor? Try the model that leans into personality.
• Prompt intentionally:
  • Ask for tone guidance (“use friendly, informal language”) if you want to dial personality up or down.
  • For critical tasks, request step‑by‑step reasoning and ask the model to show its work.
• Expect tradeoffs:
  • Richer personality can sometimes risk more controversial phrasing or speculation; cleaner responses may omit color that helps engagement.

My take

Grok winning this set isn’t an accident — it reflects a deliberate design that prioritizes human‑style conversational cues: naming trick questions, leaning into idiomatic phrasing, and using vivid details. That approach pays off in tasks where the goal is connection or storytelling.

But ChatGPT‑5.1’s steadiness is a strength, not a weakness. There are many contexts — code reviews, step‑by‑step tutorials, or corporate communications — where a measured, concise voice is preferable. The two models illustrate how “better” in AI is multidimensional: better for creativity, better for clarity, better for empathy — pick the axis that matters to you.

What to watch next

• Will developers offer hybrid flows that combine Grok‑style flair with ChatGPT’s stricter guardrails? That would be powerful.
• How will safety teams manage the balance between expressive personality and factual accuracy?
• Expect more apples‑to‑apples tests from independent outlets — these comparisons shape user adoption and product decisions.

Final thoughts

This Tom’s Guide test is a useful snapshot: Grok 4.1 crushed ChatGPT‑5.1 in this particular set of nine, especially when tone, subtext, and emotional authenticity were decisive. But the broader lesson is that the “winner” depends on what you need. The race isn’t only about raw capability anymore — it’s about the kind of conversational partner you want.

Sources

• I just tested ChatGPT‑5.1 vs. Grok 4.1 with 9 prompts — and there's a clear winner — Tom's Guide
  https://www.tomsguide.com/ai/i-just-tested-chatgpt-5-1-vs-grok-4-1-and-there's-a-clear-winner

• I just tested ChatGPT‑5.1 vs. Grok 4.1 with 9 prompts — and there's a clear winner — Yahoo (republished summary)
  https://tech.yahoo.com/ai/chatgpt/articles/just-tested-chatgpt-5-1-143010423.html