By Daniel Goetzel
In a previous post, we introduce the major opportunities coming from huge investments by tech companies into AI infrastructure, in particular OpenAI's AI Economic Zones. Here, we lay out a playbook for negotiating with OpenAI and other hyperscalers. The playbook is aimed at local and regional policymakers such as governors, mayors, county executives, and university presidents. We mainly draw from two relevant experiences. First, from the successes and missteps of the cloud computing revolution, driven by Microsoft, Amazon Web Services, and Meta over the past decade. More recently, the Biden administration's Investing in America legislative agenda, including large-scale CHIPS and Science Act innovation and economic development programs. Both provide a blueprint for making large-scale, place based economic development investments in emerging technology areas and geographic areas that have historically been left behind during tectonic technology shifts.
Structure negotiations as opportunities to unlock collective ecosystem bets vs. bilateral negotiations.
Risk: If your incentives exclusively benefit a single company, the company could leave or underperform on job creation. The deal could be viewed by legislators, activists, or other key constituents as a handout for tech titans, negotiated clandestinely in back rooms.
Opportunity: You could use the carrot of a tech company relocating to your community to make complementary ecosystem investments, such as education, commercialization and/or innovation infrastructure that outlive the deal and have spillover benefits for the community, even if the company uproots itself (say, after getting a better incentive package from your neighboring state).
Precedent: The carrot of the NSF Engines or Tech Hubs funding created new incentive structures where unlikely stakeholders teamed up on cross-institution initiatives and governments provided flexible matching funds for things that no private sector entity would fund entirely on their own (e.g. a backbone organization to support the innovation ecosystem like , , or in Baltimore). In the Tech Hubs and NSF Engines programs, we called these “but for” investments.
Virginia mastered this strategy during their negotiations with Amazon around HQ2, landing the new headquarters despite Maryland offering a package that was The Virginia Economic Development Partnership (VEDP), the state’s lead negotiator, used the negotiations as an opportunity to invest in activities that would never have made it through the legislature without the lure of landing Amazon. VEDP unlocked significant funding to create a in Northern Virginia, focused on graduating students with degrees in computer and data science. While these degrees are in technology priority areas for Amazon, they also have spillover effects that are not exclusive to Amazon. Virginia made once-in-a-generation investments in K-12, workforce programs, community colleges, and affordable housing that were only possible because of the carrot of winning Amazon. Virginia also secured Amazon's co-investment in these initiatives, including that spans the entire metro region covering MD, DC and VA.
How does this fit the OpenAI/Stargate business model? The longer it takes OpenAI to get a project through the approval process, the more money they burn. Piggybacking on a nationwide competition like NSF Engines (creating an AI track) or local competitions like the ones ongoing in , , and a stalled effort in provide opportunities for publicity, leveraged capital, and quick wins. OpenAI could offer the winner(s) a gigawatt data center plus additional ecosystem benefits that are jointly offered by the state.
Don’t confuse cloud and compute credits with meaningful financial and/or R&D co-commitments.
Risk: In the rush to announce a splashy deal with an up and coming tech company, states, municipalities, and universities accept in-kind commitments of technology (e.g. and ) that have depreciating value over time after similar in-kind technology donations flood the market. We saw this playbook from Google, AWS, Meta, and Microsoft during the and OpenAI appears to be replicating the model now as universities are desperate for computing power and GPUs. The tech companies benefit from locking researchers, startups, and city governments onto their platform at very low costs, while using researcher and city data to train their models.
Opportunity: Universities and economic development negotiators can use this technology olive branch from OpenAI as an opportunity to create meaningful, larger-scale partnership.
- White glove support for key stakeholders, helping them better understand how to make the highest use of the technology to fit their specific needs. (This appears to be already underway.)
- R&D funding for intersectional research around how new technologies can reduce data centers’ impact on the electrical grid, including funding technologies like geothermal, nuclear, and advanced materials hardware. (It appears OpenAI has already started to do this, most notably Sam Altman’s investment in but startup investments shouldn’t be conflated with meaningful research collaborations.)
- Embed OpenAI researchers in academic labs and have researchers embed at OpenAI, ideally doing these research collaboratively in-state.
- Empowering and teaching government officials to use AI to make better decisions.
- Cluster other startups and supply chain partners in the ecosystem, making buying commitments with local companies, venture investments in local startups, and incentivizing OpenAI portfolio companies to test technologies within the ecosystem.
- Create shared space in the gigawatt data centers for live experimentation with startups and academia, including hardware innovation. OpenAI, Nvidia and others could create joint R&D “data center test beds,” working with startups and local academic partners to prototype before bringing these technologies into a full-scale gigawatt environment.
Just this week, OpenAI announced a , doing many of the activities suggested above (though the details are unclear regarding how much of the $50 M commitment is cash vs. computing credits). This builds upon earlier collaboration announcements by Nvidia, OpenAI (), and Amazon ().
Precedent: A significant portion of the time spent negotiating CHIPS incentive focused on coaxing the large semiconductor companies to make meaningful R&D, university, workforce and supply chain commitments beyond technology in-kind contributions. The CHIPS team created the National Semiconductor Technology Center and two complementary funding opportunities to advance these clustering efforts, focused on and a R&D manufacturing facilities (although this program was ultimately scrapped due to overwhelming demand for other CHIPS incentives and the relocation of funding to a Department of Defense secure enclaves program). Amazon, Google, Meta and Nvidia have all launched similar efforts in partnership with academic partners in Silicon Valley and but also in places like , , and .
How does this fit the OpenAI/Stargate business model? It allows OpenAI to significantly supplement their internal R&D efforts, train models in more complex technical areas such as synthetic biology and plant genomics, source new startups for their OpenAI investment fund, create agile prototyping facilities for their portfolio companies, all while generating goodwill in these academic communities.
Develop regionally ecosystem strategies connecting rural and urban communities to create durable jobs. Think about investment holistically as opposed to individual real estate projects.
Risk: Governors, mayors, and county executives might “mortgage the farm” for short-term political wins, such as construction jobs or a ribbon cutting to announce a major project in a part of the state that has historically seen out-migration and disinvestment. The data center project could be completely disconnected from the more tech centric economic development strategies centered around university towns and larger metro areas, exacerbating a rural-urban divide.
Opportunity: There is an opportunity to transform a short-term political win into a long term bet on your region, creating a sustainable model. Data centers, given their massive space needs, often are built outside of major metro areas. Policymakers can transform these company-specific negotiations into regional-scale opportunities that create continuity and coordination between urban and rural communities. State leaders can create incentive structures that encourage local leaders to think beyond their own city or county’s borders.
Precedent: Microsoft's , over the past eight years, has invested philanthropically in exurban and rural communities where they are making major infrastructure investments, most notably data centers. They doubled down on investments that show promise such as Grand Farm, an agtech test bed in Fargo, North Dakota and Titletown in Green Bay Wisconsin. Their early investment in Fargo paid dividends when focused on agtech innovation. Their led to a robust innovation district, venture capital fund jointly managed with the Green Bay Packers football team, and a , involving close partnerships that were nurtured over time with local universities, community colleges, and manufacturers.
How does this fit the OpenAI/Stargate business model? One of the advantages of the Stargate consortium model, where Microsoft is a partner, is that the corporate partners can share costs. They are already doing so on physical and technology infrastructure but they should also think about how to do so around co-investment in ecosystem and workforce initiatives.
What's next?
You can read our follow-up post to learn about the major ongoing and upcoming AI investment collaborations and public-private partnerships that are happening already, even before the formal launch of OpenAI's AI Economic Zones.
