New Hampshire Emerging Technologies Caucus

Fostering Innovation, Collaboration, and Progress

← Back to All Events

AI as a Force Multiplier

Friday, May 29, 2026 at 10:00 AM
Presenter: Doug Alexander - SAB FOF Capital
Location: Virtual - Microsoft Teams
1516163226333.jpeg

Agenda

Speaker: Doug A. Alexander, Board Director, Amkor Technology; veteran tech CEO and investor

Intro by Rep. Terry Spahr, Grafton.

Presentation topics (30 minutes):

  • The story behind the infrastructure buildout and its supply chain
  • Perspectives on the AI revolutions’s first 1200 days and what comes next
  • Why AI is different from prior technologies - The Cognitive Revolution
  • The changing nature of the software industry and its implications for the economy, - education and geopolitics
  • The challenges and opportunities of abundance
  • What states should think about to capitalize on this technology.

Followed by member Q&A.

Open discussion as time permits.

Video

Meeting Minutes

Topic: Briefing on Artificial Intelligence — Framework, Economics, and State-Level Implications
Guest Speaker: Doug Alexander
Format: Virtual meeting
Duration: Approximately 56 minutes
Date: 29 May 2026

Attendees

  • Rep. Keith Ammon — Chair
  • Rep. Terry Spahr — introduced the speaker
  • Doug Alexander — guest speaker
  • Rep. Donald McFarlane
  • Rep. Thomas Cormen — emeritus professor of computer science (Dartmouth), co-author of Introduction to Algorithms, deputy ranking member on Science, Technology and Energy
  • Rep. John Schneller (arrived late)
  • Rep. Rick Devoid
  • Rep. Matthew Sabourin dit Choinière (arrived late)
  • John Williams

The caucus was described during the meeting as bipartisan.

Speaker Background

Terry Spahr introduced Doug Alexander, whom he met years ago in the Philadelphia area. Alexander was a co-founder of Internet Capital Group. He described himself as having 40 years in the software industry, roughly 20 years as a venture capitalist, experience building AI systems in the 1980s (first neural network around 1985), and a current board role with a semiconductor packaging firm (OSAT) that is a TSMC, Nvidia, and Apple partner building capacity in Arizona, Taiwan, and Vietnam. He stated his expertise is in AI, crypto, software businesses, and entrepreneurship.

Presentation Summary

Alexander offered a roughly 20-minute framework for interpreting AI news, followed by Q&A.

The race dynamic. He framed the current AI buildout as a "relentless race for intelligence" and a prisoner's dilemma at the company, country, and individual level: once one player accelerates, all others must follow. The strategic question driving trillions in data-center spending is who will "own you and own your company" in the AI era — a stake he sized at roughly $100 trillion, later reframed as a "six quadrillion dollar question."

Where we are on the timeline. He placed the field about 1,200 days in, analogizing the launch of GPT-3.5 to Netscape in 1994. By that mapping we are near "1998" — before the equivalents of Google, Facebook, or the iPhone — making long-range prediction difficult.

Pace of change. He argued AI is no longer a science problem but an engineering and cost problem. Costs are falling rapidly; he cited roughly 50x annual improvement in inference cost and power, compressing a Moore's-law-style gain that took 18 months into roughly three months. "Tokens are the new currency / the new oil," replacing or enhancing labor.

A new computing paradigm. Where prior computing made human cognitive work more productive, this paradigm performs cognition itself. He compared it to agriculture and the Industrial Revolution as a third major technological wave. His claim: today ~99.9% of cognitive work is done by humans; within 20–30 years that could invert to ~99.9% done by machines (with humans involved).

Collapse of the economics of expertise. Citing the "Humanity's Last Exam" benchmark (see Note 1) he noted model performance rose from roughly 3% to nearly 50% in about fourteen months, implying expert-level knowledge across many fields will become "cheap, ubiquitous, and consistent" rather than expensive and scarce. This dismantles the 200-year "expert economy" social contract and will devalue skills (he used being "good at math" as an example) that previously guaranteed middle- and upper-class status.

Read / Write / Own framing. Borrowing from Chris Dixon's book Read Write Own, he mapped AI's evolution: the "Read" age (ChatGPT-style question-and-answer, now mature), the emerging "Write"/agentic age (AI takes action on your behalf), and an "Own" layer enabled by crypto. He argued open internet protocols (HTTP, SMTP) created security problems that produced "castles" — Google, Facebook, Apple for personal data; Salesforce, Workday for corporate data — that captured enormous wealth. Crypto, properly regulated, could provide secure protocols, data ownership, and identity provenance without those gatekeepers, and could reduce the need for much of the cybersecurity industry. He praised the GENIUS Act (stablecoins) and the CLARITY Act as well-designed.

The agentic economy. He described agents that orchestrate other agents (e.g., a trip-planning agent triggering up to a thousand sub-agents), most of which will never interact with a human and will compete for compute and money. He predicted AI-built, human-free "billion dollar companies" accumulating wealth, and a coming wave of one-to-five-person micro software companies generating $1–10 million in revenue.

Science AI. He pointed to emerging "AI scientists" that propose hypotheses, design and run experiments (including robotic wet-lab work), citing a research example compressed from ~2 years and several million dollars to about four hours, and a reported ~100 AI-generated drugs approaching FDA review.

Radical abundance and distribution. He argued the internet created roughly $500 trillion in wealth, cut global poverty from ~40% to under 10%, and reduced global inequality by ~14%, and that AI could deliver "10x that" within 30 years. The open question is distribution: whether broad job creation or concentration among a handful of "trillionaires" prevails.

Advice to states. Economies that over-reward risk-taking (he contrasted the US and China against Europe) pull ahead. He urged the caucus to "skate to where the puck is going": decide what New Hampshire's role in the AI economy should be, apply AI to government to cut fraud and inefficiency (citing a hypothetical ~$200 billion/year California opportunity), and rethink K–12 and higher education. He suggested New Hampshire's strengths (proximity to MIT, Dartmouth, quality of life, attractive economic framework) could attract mobile talent and micro-businesses fleeing high-cost states like California.

Discussion and Q&A

Rep. McFarlane raised the tension between decentralized, peer-to-peer/blockchain models and the centralizing pull of sovereign (national-defense) interests and the large compute requirements of frontier models, and asked how these compete. Alexander responded that open protocols lack development incentives, that crypto is essentially a secure, agreed-upon protocol needing regulation (now beginning with stablecoins backed by Treasuries), that data centers must reside on a country's own shores for geopolitical security, and that they will migrate toward cheap/efficient energy (citing Bitcoin mining in Texas as grid-balancing).

Rep. Cormen stated he is on record opposing data centers in New Hampshire, citing energy use, cooling, noise, and the limited number of permanent jobs they create, and said he would rather see construction go toward needed housing. Alexander agreed he could not argue against that, reiterating that each state must decide and that the bigger question is positioning the state's economy rather than the binary data-center question. Rep. Ammon twice asked members to keep partisan commentary to a minimum.

Rep. McFarlane followed up on energy and taxation, noting New England's high electricity costs and distance from data-center backbones, and asked where states will diverge on taxation and policy affecting data-center placement. Alexander predicted concentration in "red states" (Texas, Florida, Tennessee, Arizona) for economic-incentive reasons, described ~$200 billion in planned Arizona capacity, and argued nuclear — including small modular reactors — is the likely long-term power answer, with the data-center industry accelerating nuclear development. He noted Arizona uses less water today than in 1957 despite far more people.

Rep. Ammon plugged a recent op-ed he wrote on pairing small modular / advanced nuclear reactors with data-center demand, calling the pairing "peanut butter and jelly." Alexander agreed building nuclear adjacent to multi-billion-dollar plants would be sensible, and noted energy access is widely viewed as the ultimate constraint on the AI buildout.

Closing

Rep. Ammon wrapped the meeting near the hour mark, thanking Alexander and noting the conversation would continue online. Alexander offered to return and to be reached by email, and commended the caucus's work, saying he hoped every state was doing the same. Rep. Cormen gave a brief background statement at the end (Dartmouth, Introduction to Algorithms, Science/Technology/Energy committee). Terry Spahr thanked the speaker.

Action Items / Follow-Ups

  • Continue the discussion online (per Rep. Ammon).
  • Members may email Doug Alexander with further questions; he is open to a repeat session.
  • Caucus to consider New Hampshire's strategic positioning in the AI economy (talent attraction, micro-business climate, education, energy/nuclear policy).

Notes

  1. Humanity's Last Exam. For the record: the benchmark was created by Dan Hendrycks (Center for AI Safety) together with Scale AI, reportedly inspired by a conversation with Elon Musk who felt existing benchmarks were too easy — it was not co-created by Musk. The published benchmark contains about 2,500 questions across roughly eight broad subject areas (the speaker's "3,000 questions / 100 subjects" was imprecise), contributed by nearly 1,000 subject experts from more than 500 institutions across 50 countries. The score progression cited by the speaker (low single digits at the January 2025 launch rising toward ~50%) is directionally accurate.

Sources

The following were used to verify factual references in these minutes:

AI as a Force Multiplier | NH Emerging Technologies Caucus