Ocean data centers, lawyer hiring, Super Bowl trades
Being there when the tide goes out, inputs matter, blockbuster events
Bobbing for data centers
Panthalassa, a startup building self-contained floating data centers for use in the ocean, is the product of ten years of smart applications of simple concepts mixed with great science and engineering.
First, they started with a problem. “It’s going to be too expensive to mine bitcoin in the future!” This was in 2016. OK, then comes the smarts to start executing over the next ten years. Set out to build a floating lollipop the size of the Empire State Building, push it out into the deep ocean where it can bob up and down with the waves, and have it generate power to mine the bitcoin all in a self-contained unit that sends data out via satellite. Cool.
Suspend disbelief for a moment and assume that would work.
This is the application of right tool for the job taken to creative levels. For a long time, we have known that you can harness energy from waves. We also know that it is very difficult and expensive to transfer that energy back to land where it could power homes and the grid. So, what if you found something to do with that energy out amidst the waves and never bothered to bring it back to shore, cutting tons of complications and costs? They realized that compute and data can travel but it’s better if energy doesn’t also have to travel.
OK, now we are out there—and the giant lollipops could be 200 or 1000 miles off-shore in the deep ocean. What could we do with the energy out there? Well in 2016, bitcoin was trading under $1,000. So, early days. Undeterred by the potential lack of demand, the team realized that skating to where the puck is going is very smart for a team building prototypes that would take years to fully test, never mind commercialize.
Now, of course, the best laid plans don’t always come to fruition if you are banking on other developments like bitcoin mining’s demand, so it’s best if you have suspenders and a belt. When working on things that take a long time, it’s good to have optionality built in. For Panthalassa, this meant that they had a few thoughts on the roadmap for what they could make power for if bitcoin demand dropped off.
They thought maybe AI inference could be a thing (this worked out well). In 2016, AI inference and ocean data centers weren’t completely crazy. Microsoft had a program, Project Natick, to build undersea data centers that led to deployed prototypes starting in 2015. Microsoft’s undersea approach wasn’t the right one, but it did validate underlying demand in the early days.
They also figured that eventually hydrogen or other fuels could be made in the same container if there wasn’t enough AI demand. Bitcoin mining, data center AI inference, and hydrogen production gave them three different options for what to do with the power at sea and none of them required sending any power back to the mainland if they could get it working.
Well, it turns out it is now well on its way to working.1
Panthalassa, which is the name of the giant world ocean that surrounded Pangaea, is progressing its vision of useful, cheap, clean ocean power. They figured out how to make it work without energy coming back to shore, no tethering or anchoring of the giant floating lollipops needed, self-contained and can propel itself around the ocean looking for great wave spots like a Roomba. The cold ocean water can be used to cool and power (via turbines) the onboard data center. It never gets near the ocean floor, so there isn’t nearly the environmental impact of ocean mining or anything on the seabed.
They think they could potentially get the energy costs out in the ocean down to 2 cents per kWh. Today, on-shore energy costs for data centers are typically 6 to 20 cents per kWh.
They announced yesterday that Peter Thiel is leading a $140M investment into the company at around a $1B valuation. Super Micro Computer, one of the world’s largest producers of high performance servers, also invested into the round. There are still some risks on the table to get this scaling commercially but the path is impressive in its simplicity: start with the problem, find the right tool for the job, and if it’s going to be a long journey, bring suspenders and a belt.
Trading Super Bowls for F1 races
Miami has chosen to give up future Super Bowls. Not competing for them,2 but hosting them. At a time when seemingly everyone is ready to accommodate the NFL in order to host the Super Bowl, Miami has chosen to bow out from vying completely. Miami last hosted the Super Bowl in 2020 and has hosted 11 of them in total,3 roughly one every six years.
Hard Rock Stadium and the surrounding area has been built up to accommodate other major events and those changes around the stadium would impede the hospitality areas that the NFL requires from hosts of the Super Bowl. Miami did meet the NFL’s standards in 2020, but then chose to make improvements that slid it out of compliance with the league’s requirements.
What did Miami get in return for the changes? It hosted the first Miami F1 event in 2022. The event is now held annually and brings approximately 275,000 visitors for the three-day event.
It has also configured the area to suit the Miami Open for tennis each year, which up until 2019 was held elsewhere in the city. Stephen Ross, the Dolphins owner, cites the work they did to get the Miami Open to the stadium area as crucial for securing the F1 race. F1 was initially skeptical of having the race near the stadium (they wanted downtown Miami) but came around after seeing the success of the tennis event.
Hard Rock Stadium has been seeking other major tentpole events in the calendar. This year, it has hosted the NCAA Men’s Football National Championship which saw a huge influx of Indiana Hoosier fans. There are going to be seven FIFA Men’s World Cup games at the stadium this summer. Prior to the last Super Bowl in Miami, the stadium was seeing 25 events a year. Now it hosts around 60 each year.
We can look at the ledger here and see that Dolphins’ ownership weighed a once-every-six-years Super Bowl vs likely losing the Super Bowl but building out to hold F1 and secure tennis. The success of F1 in Miami and its predictable annual cadence presents a lower risk profile in comparison to the irregular, blockbuster Super Bowl. The cash flow is smoother and there is a tentpole and schedule that can be built around.
CNBC's Andrew Ross Sorkin recently asked the Dolphins’ owner: “I think you made more money on F1 than the Dolphins. Am I wrong?" And Stephen Ross replied: "F1 has been great […] We get more attendance for F1 races for three days than the entire (Dolphins) season tickets that we sold."
Economic impact studies of sporting events can be notoriously unreliable. Here we see an owner making a clear choice and after watching it play out for a five-year cycle continue to make the same choice. If the NFL loosens its requirements or works with the Dolphins to facilitate a 2030 Super Bowl in Miami, that would be nice upside.
Miami saw an opportunity with tennis and F1 and went for it. Now, it is established on the calendar with F1 and racing fans. Had the Dolphins stuck with the Super Bowl, that annual race might just as easily be elsewhere. The time to eat is when hors d’oeuvres are being passed.
Which way, young lawyer?
The two most popular categories of AI-legal startups are the AI-for-lawyers (eg Harvey, Legora) and the AI-native law firm (Crosby). Who and what they are hiring for is instructive in thinking about how the legal industry talent pipeline will shape up.
Lots of air has been exhaled on X (formerly Twitter) in recent days about Harvey and Legora. Crosby announced their Series B earlier this month. Valuations across the space range from $500M to $11B. Discussions have centered around who is going to win the Legora vs Harvey race and occasionally will drift towards whether there is a place for Crosby as well.
Not discussed is that inputs shape these businesses and the two different business models have very different inputs.
AI-for-lawyers (Harvey, Legora) are built with a few key building blocks. They sell to existing law firms that already are well entrenched with thousands of lawyers, armies of associates, researchers and established rosters of clients that they advise and do work for across a huge range of work, from litigation to corporate law. They take leading AI models, enhance for lawyers’ usage however possible, and package it into a law firm friendly implementation and sales cycle.
AI-native law firms (Crosby, as an example) start from the premise of a reimagined law firm—as if aliens had landed and brought with them AI technology and decided to use the AI to build a new law firm from the ground up. Here there is a marriage of AI and lawyers working together to have highly efficient, highly effective lawyers who are (in theory) able to outcompete old-school law firms on price, speed and quality of work. They sell to up-and-coming startups and businesses who need legal services. They don’t sell at all to other law firms. The AI is all behind the scenes where AI and lawyers work symbiotically to deliver for the client who doesn’t really care how the sausage is made.
Who you sell to for this type of technical product, which goes well beyond comparing drafts and marking changes, makes a world of difference in who you need to hire. AI-for-lawyers companies have gone out and hired a lot of young lawyers to go into law firms and help implement the AI. The forward deployed lawyer is a thing. The lawyer portion of that does a lot of work. Law firms can be idiosyncratic and some are resistant to change. As an indicator, Harvey recently stated that they see active users spend 30 mins per day on average in Harvey—which is still quite low considering the hours and hours of researching and drafting that lawyers do on a daily basis.
The forward deployed lawyer aims to move that number up considerably by embedding in the client’s organization. The upside here is that Harvey/Legora get really embedded in these workflows and lawyers become comfortable and familiar. The downside is that they need to pay lawyers to do this work and that means salaries high enough to attract lawyers and share based compensation (which is real). The early tenure lawyer who joins has to consider that they aren’t doing full law work anymore and needs to weigh the impact on their career. For those looking to transition out of law and into tech, this is a plus.
AI-native firms also recruit lawyers but not for sales roles. The lawyers work in a well-appointed boiler room and they practice law.4 They are paid like lawyers and their career trajectory can continue to be law focused. The AI-native firms can hire sales talent to do sales and to the extent there could be some implementation work, it can be handled by non-lawyers. The lawyers are part of the marginal cost of the business and are firmly outside sales and marketing, unlike some of their fellow lawyers at AI-for-lawyers where they can be part of sales and marketing.
Even though both types of businesses recruit lawyers, there is variation in looking for career switchers vs die-hards who want to take down more contracts in a day than before.
AI-native and AI-for-lawyers might not end up being competitive at all, outside of the inputs. They sell to different customers completely. Sequoia is an investor in both Crosby and Harvey. As I wrote about the Finix example, they work to avoid conflicting investments. Your inputs shape the business.
Like these essays? I write a trio of essays on wide ranging topics delivered four times a week. Subscribe and forward it to your friends. Have feedback, thoughts to share or comments? Comment or reply.
Debatable in 2026
Currently tied with New Orleans for the most. The Super Bowl is an entertainment product, after all.
Probably not a boiler room at all. Probably a very enjoyable environment for young lawyers.