TechCrunch Disrupt 2025 lands at Moscone West in San Francisco on October 27�29 with a star-packed speaker list and 10,000-plus founders and VCs expected? But behind the demos and pitch sessions, the real conversations in the corridors are likely to center on three hard constraints shaping AI�s next year: electricity, regulation, and public trust?
Compute�s new moat: a power plant
AI�s growth story is increasingly an energy story? In West Texas, Poolside�s Horizon complex spans 500 acres and targets an eye-popping 2 gigawatts of compute, according to TechCrunch reporting? OpenAI�s �Stargate� project near Abilene is slated for roughly 900 megawatts across eight buildings? And Meta�s planned $10 billion facility in Louisiana is paired with 2?3 gigawatts of new natural gas generation from utility Entergy dedicated to its needs? That�s not future hype; it�s concrete, steel, and long-term power contracts?
Why the pivot to self-provisioned fossil power rather than waiting on the grid? Reliability and speed? �We�re burning gas to run this data center,� OpenAI�s Sam Altman acknowledged, while the company�s global affairs chief Chris Lehane argued the scale-up could �re-industrialize� the U?S? if done right? Critics counter that these choices risk locking in emissions and water use: a Duke University study cited by TechCrunch found utilities operate at about 53% of available capacity on average, and shaving peak demand by half could free up 76 gigawatts�suggesting smarter demand management might beat building new gas plants? Local residents are also feeling the strain; one Abilene-area homeowner described construction noise and �bright lights� disrupting the quiet life she�d sought decades earlier?
For founders and CTOs heading to Disrupt, this isn�t abstract? The questions now sound more like infrastructure finance than app development? How will you secure five to 15 years of predictable power? Will you buy on-grid with demand response, co-locate with generation, or sign behind-the-meter deals? And how do you square that with sustainability commitments customers increasingly require?
Safety vs? speed is no longer a Twitter fight�it�s a compliance plan
The broader policy climate may be as consequential as any product roadmap? California�s SB 53, signed last month, adds safety reporting duties for large AI providers? That�s in the same week Silicon Valley heavyweights renewed their feud with AI-safety advocates: White House AI czar David Sacks accused Anthropic of a �regulatory capture strategy,� while OpenAI�s Jason Kwon defended subpoenas to nonprofits critical of the company, citing transparency concerns? Internally, even OpenAI alignment leaders flagged discomfort with the optics?
What�s the business takeaway? If you build or integrate frontier models, expect more disclosures on evaluations, incident reporting, and model update practices? Early movers who can evidence risk controls without stifling pace will be more credible in enterprise sales cycles�and with regulators? Roughly half of Americans say they�re more concerned than excited about AI, according to survey data referenced by TechCrunch? That sentiment isn�t solved by a better demo; it�s solved by verifiable processes?
Deepfakes hit official politics�brand safety follows
AI-generated misinformation crossed a new line when Senate Republicans posted a deepfake of Chuck Schumer celebrating a government shutdown on X? The video carried an AI watermark, yet the platform neither removed it nor applied a warning label despite rules against harmful manipulated media? With up to 28 states enacting limits on political deepfakes�and some banning those intended to sway elections�enforcement will vary by jurisdiction and platform?
For companies, the implication is clear: election-year risk will spill into brand safety and customer comms? Content provenance tools and takedown playbooks need to be paired with model-driven detection pipelines? Don�t count on platform enforcement to catch everything�and be ready to prove your own content is authentic?
What to watch on stage�and what to ask in the hallways
Disrupt�s lineup features operators who will be central to these debates, including Microsoft CTO Kevin Scott, Netflix CTO Elizabeth Stone, Waymo co-CEO Tekedra Mawakana, and investor Vinod Khosla? The Startup Battlefield will spotlight early-stage bets where these structural questions can make or break scaling?
- Power procurement as strategy: Are startups negotiating multi-year power deals, or partnering with cloud providers that can credibly guarantee capacity without offloading emissions?
- Safety by design: How are teams documenting evaluations, fine-tuning processes, and incident response to satisfy SB 53-style reporting while maintaining release cadence?
- Trust in the loop: What content provenance, watermarking, or authenticity standards are product teams adopting to reduce deepfake and hallucination risk for customers?
- Cost of capital for compute: If the cycle shifts toward private energy-backed data centers, how will financing terms, PPAs, and utility relationships reshape who wins in AI infra?
In short, Disrupt 2025 will likely be less about the next clever prompt and more about permits, power purchase agreements, and proof? The companies that thread those needles�balancing scale with stewardship and speed with safety�will set the tone for 2026?