OpenAI transformed ChatGPT into a “chat OS,” IBM Ventures drew 1,700 to its waitlist, and 58 startups pitched in rapid-fire format during Silicon Valley’s largest-ever decentralized tech festival. From October 6-12, San Francisco became a city-wide technology showcase with over 1,000 independent events and an estimated 15,000 attendees, marking a pivotal shift from AI experimentation to enterprise-scale deployment. The week revealed an industry focused on autonomous AI agents, corporate venture capital’s resurgence, and practical monetization over hype. Organized by Andreessen Horowitz, the festival spanned dozens of SF neighborhoods with everything from OpenAI’s DevDay product launches to exclusive founder dinners at Michelin-starred restaurants, establishing SF Tech Week as the premier gathering for the global tech ecosystem.
OpenAI stole the show with platform play
OpenAI’s third annual DevDay on October 6 at Fort Mason opened SF Tech Week with a strategic pivot from AI model provider to complete platform ecosystem. Before 1,500 developers, the company announced AgentKit, a production-grade toolkit for building, deploying, and optimizing AI agents featuring a visual canvas with drag-and-drop logic flows, connector registry for third-party integrations, and built-in observability with tracing and performance metrics. The announcement signaled OpenAI’s ambitions to compete with traditional software platforms rather than just supply models.
The company unveiled Apps in ChatGPT with Apps SDK, transforming ChatGPT into what executives called a “chat OS” where third-party apps run directly inside chat interfaces. Launch partners including Spotify, Canva, and Zillow can now create interactive experiences through embedded cards, with commerce integration planned for in-chat purchasing. Built on the Model Context Protocol, the platform offers standardized data access for developers building context-aware tools.
API access opened for GPT-5 Pro (high-accuracy model for enterprise workloads) and Sora 2 (advanced video generation with aspect-ratio controls and sound generation, priced per-second). Platform metrics revealed impressive scale: 800 million weekly ChatGPT users, 4 million developers building on OpenAI, and 6 billion tokens processed per minute. Sam Altman delivered opening remarks, while former Apple design chief Jony Ive joined for a fireside chat on “The Craft of Building in the Age of AI,” drawing significant buzz around potential future collaborations.
Corporate venture capital made its comeback
The week’s hottest ticket wasn’t a product launch or keynote—it was “Scaling Over Cocktails,” IBM Ventures and a16z SPEEDRUN’s networking event that drew 1,700 people to its waitlist. This extraordinary demand illustrated what Emily Fontaine, IBM’s Global Head of Venture Capital, described as a fundamental shift: “The tide has changed for corporate ventures. Founders want corporate investments on their cap table at an earlier stage.”
IBM’s $500 million Enterprise AI Fund made aggressive moves during the week, with Fontaine emphasizing that corporate VCs now offer operational support founders desperately need. “The most difficult thing founders have is executing deals and getting them through procurement for enterprise clients,” she explained. IBM’s network can “speed up deals” through established client relationships—more valuable than just capital. Corporate and corporate venture capital now accounts for 35% of total deal value, the highest proportion since 2019.
Recent IBM portfolio investments showcased the fund’s focus: Not Diamond (dynamic AI model routing), HiddenLayer (ML security), Hugging Face ($235M Series D participant), and Unstructured ($40M Series B for enterprise data preprocessing). Fontaine dismissed AI bubble concerns: “There’s a ton of capital on the street. Everyone’s got money to spend. I don’t think there’s an AI bubble—the best safeguard is rigorous diligence.”
SPEEDRUN Demo Day showcased AI-native startups
Tuesday, October 7 brought a16z’s SPEEDRUN Demo Day (SR005) to SFJAZZ Theater in SOMA, where 58 pre-seed and seed-stage startups pitched to 500+ investors in a rapid-fire format. Each founder had exactly two minutes on stage—no introductions, no applause—creating what attendees described as “watching the future on fast-forward.” The program invested up to $1 million per company plus $5 million in credits from AWS, OpenAI, and NVIDIA.
Notable companies included Zero Billion (AI-native TV platform, $2M revenue), Shockwave (AI gaming platform with $2.3M revenue and 180K daily users), Azimov (engine for AI applications, $2.5M ARR), Presia (AI agents for management consulting, $270K ARR in just 3 months), and Margin (first AI-powered credit card). The cohort’s emphasis on revenue traction marked a departure from earlier funding cycles—these weren’t concept-stage companies but startups already generating meaningful revenue.
The 58 companies spanned AI agents, enterprise software, gaming, fintech, and hardware. Geographic diversity included companies from SF, NYC, LA, London, Israel, and Toronto. The program’s less-than-1% acceptance rate and 12-week intensive format emphasized velocity, with one founder noting: “Speedrun reinforced in us the importance of velocity—shipping and learning quickly.” 500 Global also held their Batch 36 Demo Day that afternoon in North Beach, investing $150,000 per company at 6% equity.
Agentic AI dominated the conversation
If 2024 was about foundation models and generative AI, 2025 was unequivocally about agentic AI—autonomous systems that can reason, plan, and act. Multiple dedicated sessions explored this theme: “LLM Post-Training in the Agentic Era” (Dedalus Labs, Anthropic, OpenAI), UC Santa Cruz’s “AI Frontier: Data, Agents & Robots” (200+ attendees), and AI Agent Security Summit addressing new cybersecurity concerns around autonomous systems.
The shift from pilots to production emerged as a critical theme. The panel “From Pilots to Scale: What Founders Need to Know About Enterprise AI Adoption” featured Puzzle, Rippling, and Perplexity discussing practical challenges: navigating enterprise procurement, achieving SOC 2 compliance, building trust with corporate buyers, and demonstrating ROI beyond proof-of-concept. UC Santa Cruz Professor Ian Lane noted: “There’s a lot of thought that software development is almost a solved problem, and it was really great to hear from people running companies that it is not, and there’s still a long way to go.”
Emily Fontaine identified specific emerging areas gaining traction: “vibe coding” (AI-assisted development), multimodal strategies (AI handling multiple data types), and reinforcement learning. Not Diamond raised funding for intelligent AI model routing, with co-founder Tomás Hernando Kofman explaining: “The more we can leverage the strengths and weaknesses of many diverse models instead of relying on a single monolith, the more we move towards a safer, more performant future for AI.”
Major AI companies hosted exclusive gatherings
Beyond OpenAI’s DevDay, other AI leaders created intimate networking opportunities. Anthropic held an invite-only AI Founder Salon on October 7, featuring deep dives into Claude’s latest features led by co-founder Ben Mann, with focus on interpretability research and scaling strategies. Mistral AI hosted an AI Leaders Dinner (invite-only) on October 9, while ElevenLabs (voice AI) organized multiple events including a private VIP dinner in Presidio Heights and networking at the SF Giants game on October 8.
Anthropic also participated in the “LLM Post-Training in the Agentic Era” technical session with Dedalus Labs and OpenAI, attracting developers interested in advanced training methodologies. The competitive dynamic between OpenAI, Anthropic, Google’s Gemini, and other players shaped discussions throughout the week, with companies positioning themselves as the platform for the agentic era.
Vercel and NVIDIA compressed a hackathon into 2 hours
One of the week’s most buzz-worthy events was Vercel × NVIDIA’s “World’s Shortest Hackathon” on October 9—a full hackathon compressed into just 2 hours. Limited to 50 teams of 2 (100 participants total), the event demonstrated how AI-powered development tools have accelerated software creation to unprecedented speeds. Participants used AI prompts and coding assistants like Vercel V0, Cursor, Bolt, and Replit AI to build complete applications in record time.
The winning team received a GeForce RTX 5080 GPU signed by NVIDIA CEO Jensen Huang plus Vercel credits. Judging criteria emphasized real-world novel applications, highly leveraged generative AI tooling, and overall “vibes.” The event embodied the “100X engineer” concept—developers becoming exponentially more productive through AI assistance. The 2-hour format, previously held at NY Tech Week in June and GTC 2025, symbolized the velocity driving the industry.
California passed first-in-nation AI regulation
Regulatory discussions permeated the week following California Governor Newsom’s signing of SB 53 (Transparency in Frontier Artificial Intelligence Act) in September 2025. This first-in-nation law established a regulatory framework for advanced AI systems requiring catastrophic risk assessments from developers, reporting to California’s Governor’s Office of Emergency Services, whistleblower protections, and civil penalties up to $1M per violation for noncompliance.
Fenwick, the event’s exclusive legal sponsor, hosted “Law Meets Code” sessions with a16z and Databricks titled: “Upscale responsibly and stay ahead of the competition with critical strategies for navigating the shifting AI regulatory landscape.” Topics included AI safety, ethical standards, and governance frameworks. The regulatory focus contrasted with the Trump Administration’s “AI Action Plan” calling for a moratorium on state AI regulation—which the Senate voted 99-1 to strike. California’s framework may serve as a model for other jurisdictions emphasizing transparency over prohibition.
The party scene rivaled the programming
SF Tech Week’s decentralized format created a festival atmosphere with parties and networking events across dozens of venues. Thursday, October 9 emerged as peak party night with Atlassian’s Startup Unleash Party (4:30 PM, FiDi), Valon Nights rooftop cocktails (6:00 PM, SOMA), and the SF Tech Week Afrobeats Party (300+ tech professionals, hosted by TravelBayOnline) offering “where tech meets vibes, fun, and culture.”
Exclusive gatherings included OpenAI’s Founder Reception (invite-only, Mission Bay), Mistral AI’s Leaders Dinner (invite-only, Downtown SF), and Private VIP Dinner with ElevenLabs (Presidio Heights). High-end experiences featured Truffle Security Dinner at Quince (upscale restaurant, Jackson Square) and the Emerging Managers + Founding GPs Omakase (Japanese omakase dinner, SOMA). Sports experiences brought unique networking: ElevenLabs at SF Giants game (corporate box, South Beach) and Golden State Warriors courtside with Oleum AI.
Early-morning wellness events included Perkins Coie’s Surf Meetup for Founders (7:30 AM, Ocean Beach), Pigment’s Women in Finance Breakfast & Puppy Yoga (Dogpatch), and various game nights. Google Cloud & Sequoia at The Stadium (October 7, South Beach) and J.P. Morgan’s Taste of Innovation (invite-only, FiDi) showcased corporate investment in high-production events.
Notable speakers and thought leaders participated
Beyond keynotes, prominent voices shaped discussions throughout the week. Andrew Chen (General Partner, a16z) maintained a prominent presence at SPEEDRUN and growth strategy sessions. May Habib (CEO, Writer.com) spoke on “AI in Enterprise: Practical Applications and Future Opportunities,” sharing real-world scaling experiences. Shannon Miller (Senior Vice President, IBM) discussed “Enterprise AI Transformation and Implementation.”
Baron Davis (former NBA All-Star turned tech investor) brought perspectives on transitioning from sports to technology investing. Tara Viswanathan (CEO, Rupa Health) delivered health tech innovation talks at Hustle Fund events. Aravind Srinivas (CEO, Perplexity AI) participated in programming, with his company processing 780M queries in May 2025 with 20%+ month-over-month growth.
Fireside chats featured Sam Shank (HotelTonight co-founder) and founders from Venmo and Cal.com. Jony Ive’s appearance at OpenAI DevDay generated particular buzz, with speculation about potential AI hardware collaborations. The diversity of speakers—from hardcore technologists to enterprise executives to celebrity investors—reflected the week’s broad appeal.
Major themes emerged from hundreds of sessions
From hype to revenue dominated strategic discussions. Multiple sessions focused on “Tactical advice on GTM & Growth,” with emphasis on actual monetization strategies rather than just technological capabilities. Attendees noted meeting founders “already doing a few million ARR at Seed stage”—a stark contrast to earlier funding cycles prioritizing growth over profitability. The focus on capital efficiency and extending runway reflected market maturation.
AI × Creativity sessions explored how AI transforms creative industries. Gold House Ventures, Mantis VC, OpenAI, and a16z collaborated on “AI x Creativity: Future of Creation” (October 8), examining audio generation, multimodal creativity tools, and new paradigms for content creation. ElevenLabs’ multiple events highlighted voice AI’s growing importance.
Global expansion and distributed work featured prominently with Deel’s involvement signaling distributed work infrastructure importance. Multiple sessions covered UK, LATAM, Irish, and South Asian expansion playbooks, immigration strategies, and cross-border operations. The Information & NYSE panel “IPOs, Exits & What’s Next” (October 9, SOMA) discussed exit strategies and public market conditions, with nearly two-thirds of VC fund managers expecting exits to increase over the next 12 months.
Funding landscape showed paradoxical signals
The broader funding context revealed market complexity. Q3 2025 global VC totaled $97 billion (up 38% year-over-year from $70B in Q3 2024), with AI sector capturing $45 billion (46% of all VC funding). Mega-rounds ($500M+) represented 18 deals and 33% of all funding, including Anthropic ($13B), xAI ($5.3B), Mistral AI ($2B), and Cerebras Systems ($1.1B at $8.1B valuation).
Yet paradoxically, venture firms themselves were on track to raise the least capital since 2017. As PitchBook’s Kyle Stanford noted, this combination gives a “holistic view of just how strange this market is.” Exits increased but at valuations far below 2020-2021 boom levels. Seven Bay Area companies in Q3’s top 10 deals garnered $28.8B—36% of all US VC in the period—demonstrating SF’s continued dominance despite talk of geographic dispersion.
Through SPEEDRUN and 500 Global demo days alone, SF Tech Week represented minimum $58M+ in direct capital deployed, with estimated deal flow of $200-500M that will materialize in announced rounds over Q4 2025 and Q1 2026 based on historical patterns.
Web3 made limited but strategic appearances
While AI dominated, Web3 maintained presence with MiniDoge’s ecosystem launch headlining the Orbis86 Venture Summit and sponsoring BuildETH 2025 Conference (all-day Ethereum and DeFi event). MiniDoge showcased its technical roadmap for “empowering Web3” and “reshaping the Meme Landscape,” marking a pivotal step toward Western market expansion after establishing presence in Dubai, Malaysia, South Korea, Hong Kong, and Singapore.
BuildETH 2025 focused on Ethereum and decentralized finance, with MiniDoge serving as title sponsor. However, Web3’s reduced prominence compared to previous years reflected the industry’s shift toward AI, though blockchain advocates maintained that foundational infrastructure work continues despite lower visibility.
Event sponsors shaped the week’s programming
Andreessen Horowitz (a16z) served as presenting sponsor and primary organizer, leveraging SF Tech Week to showcase its SPEEDRUN accelerator and portfolio companies. Platinum sponsors included Workday (AI platform for HR) and a16z speedrun. Exclusive legal sponsor Fenwick hosted multiple regulatory and strategic sessions.
Major corporate participants included IBM (multiple events, $500M fund), AWS (Cyber60 celebration with Lightspeed), Google Cloud (Stadium event with Sequoia), Microsoft (Founder × Funder Dinner via Microsoft for Startups), NVIDIA (hackathon with Vercel), Atlassian (Startup Unleash Party), Databricks (regulatory discussions), and HSBC Innovation Banking (fintech events).
PwC Emerging Company Solutions, Hustle Fund, 500 Global, Gold House Ventures, Mantis VC, Lightspeed Venture Partners, and Silicon Valley Bank all hosted programming. Legal firms Morrison Foerster and professional services firms like Dataiku contributed to ecosystem support. The diversity of sponsors—from VCs to tech giants to law firms to banks—demonstrated SF Tech Week’s position as a truly ecosystem-wide gathering.
The decentralized model proved its value
SF Tech Week 2025’s format—over 1,000 independent events across dozens of SF neighborhoods—contrasted sharply with traditional conferences. Events concentrated in SOMA (near a16z offices), Financial District, South Beach, Mission Bay, and spread to Jackson Square, Dogpatch, Presidio Heights, and North Beach. This geographic distribution created “serendipitous encounters” as attendees moved between venues.
The model maximized networking density. Rather than being “funneled into formal corridors,” attendees crossed paths “at dinners, co-working spaces, and local events.” The mix of open-registration events and invite-only gatherings created tiered access, with top AI companies and VCs hosting exclusive receptions while maintaining broader community programming. The format generated FOMO—with multiple simultaneous high-value events, attendees constantly faced trade-offs about where to be.
Attendee testimonials reflected enthusiasm: “Day 1 at #SFTechWeek was incredible! Met so many inspiring founders” and “No better way to describe my first #SFTechweek experience except that it was everything and more than what I expected it to be.” The scale—15,000+ attendees across the week—established SF Tech Week as arguably the premier tech gathering globally, rivaling or exceeding traditional conferences in reach and impact.
Conclusion: A pivot point for the industry
SF Tech Week 2025 marked the tech industry’s maturation from AI experimentation to enterprise deployment, from funding euphoria to disciplined capital allocation, and from unregulated innovation to emerging governance frameworks. The agentic AI revolution emerged as the consensus bet for the next wave of value creation, with OpenAI’s platform play, SPEEDRUN’s agent-focused startups, and countless panels all pointing toward autonomous systems as the future.
Corporate venture capital’s resurgence—exemplified by IBM’s $500M fund and 1,700-person waitlist—signaled that founders now seek strategic value beyond capital. The week’s emphasis on practical monetization, enterprise adoption, and regulatory compliance contrasted with previous cycles’ growth-at-all-costs mentality. California’s SB 53 regulation demonstrated that governance frameworks are arriving whether industry prefers them or not.
The decentralized format succeeded in creating a week-long immersive experience that transcended traditional conference boundaries. From OpenAI’s DevDay announcements to exclusive dinners at Michelin-starred restaurants, from 2-hour hackathons to surf meetups, from 58 rapid-fire pitches to rooftop cocktails overlooking the Bay—SF Tech Week 2025 established itself as the global tech community’s most important annual gathering. The question heading into 2026 isn’t whether SF Tech Week will continue, but whether any other event can match its scale, density, and influence.