How an AI expert is investing given this month’s blockbuster AI news
It was a big week for AI news in Massachusetts and across the country. The Kraft Group announced a plan to build a huge campus with up to 10 data centers in Sutton. Teachers, parents, and students protested outside Google's Cambridge office over the company's decision to add AI features to its education products. NVIDIA, a company with a small research lab in Massachusetts focused on quantum computing, also saw its CEO resign after reports about two alleged affairs outside his marriage.
Meanwhile, Anthropic, a California company that is growing its presence in Cambridge, is expected to go public this week.
Doug Levin has been following much of this activity. He writes a Substack blog called “Lessons from a Startup Life.” He also helped create an AI course at Harvard Business School and is a fellow at Stanford University. I spoke with Levin on Monday. (This interview has been edited to make it easier to read and understand.)
Scott Kirsner: Anthropic is going public this week. Are you going to buy its stock? Why or why not?
Doug Levin: Anthropic's estimated value is $1.4 trillion, based on a share price of $70 to $78. That price already assumes the company will make a lot of money in the future, so I will not buy a lot of its stock. If the company had a more reasonable value, I would buy more than I plan to buy now. I also think this could be a very successful IPO, with the share price rising well above $100. That would give the company a value of about $2 trillion. But that is very expensive.
Kirsner: So it sounds like you plan to buy a little, just to see what happens.
Levin: Exactly.
Kirsner: Let's talk about the AI race between the United States and China. In some ways, the Trump administration seems to be saying, “Let's put no limits on our AI industry because we have to beat China.” Do you agree? Should there be no limits because this is the new space race?
Levin: One thing you may have missed this morning was a report in The Washington Post that Trump had built up a $62 million investment in NVIDIA. When I see decisions coming from the Trump administration, I immediately think they are being made because of corruption and investments in these companies. The administration does not have a fair view of the situation. It does not put America first. It puts its own investments first. That is my first point.
My second point is that one of the biggest fools in the industry, David Sacks, the president's science and technology adviser, helped set the administration's AI strategy in motion. And it is a mess.
AI policy is being made not by the federal government or Congress, but by 50 different states and local governments. This is happening because cities, towns, and counties are worried about AI surveillance, data privacy, and AI tools that may go out of control, such as AI agents. The federal government has failed the public. Congress has failed. The White House has failed.

Kirsner: If you were in China's position, would you think it had the same attitude as Donald Trump? Would China say, “Let's do as much as possible, as fast as possible, with no limits on AI”?
Levin: No. It is almost the opposite. China has a long-term plan. It wants to build open-source AI models, weaken American companies such as Anthropic and OpenAI, and help as many Chinese people as possible build data centers wherever they can, as close to water sources as possible. I think about 20% of Chinese data centers use renewable energy. China has a better plan for energy use than we do. That is because the Trump administration is too closely tied to the coal industry and its allies in the Gulf region.
Kirsner: Since we are already talking about data centers, what did you think last week when the Kraft family announced plans to build a huge data center in Sutton, Massachusetts? It was just the week after the midterm elections, when data centers were a major issue in many races. Then the Krafts came along and said, “OK, everyone, here is our plan.”
Levin: They are probably looking at the business model and thinking that data centers will be a very good business in five or 10 years. Building them costs a lot of money in the short term, but they are probably thinking about the long term. They are smart people who understand the opportunity.
I think we have too many rules. We already have a responsible plan for building data centers, issued by Governor Maura Healey in June. We also have state orders that add rules for getting building permits. This is typical of Massachusetts. People here think having too many rules is a good thing. But that is not always good for new ideas and businesses in Massachusetts.
Does Massachusetts want to be a leader in the future? Whether we like it or not, we will need data centers here. We just need to choose their locations carefully so we can make good use of them.
Kirsner: You said we have too many rules here. There is some AI legislation being considered. It could create the toughest AI oversight rules of any state. Independent experts would examine the most advanced AI models and ask whether they could cause a disaster. It seems that both OpenAI and Anthropic support this idea. Do you think this is an example of why we should not have the toughest AI safety rules, or could it be a good example for other states?
Levin: I think the best way to handle this is to make the boards of directors of companies that develop the most advanced AI models responsible. If their own jobs and money are at risk, they will put safety measures in place.
Kirsner: Do you mean making companies responsible if bad people use their technology?
Levin: The boards of directors at OpenAI, Anthropic, and other companies that build the most advanced AI models, as well as other AI developers, would approve company rules. These rules would provide safety limits, monitoring, and management of AI products. That way, if an AI system went out of control, people could trace the problem back to the company.
Kirsner: Sorry, in your plan, would board members face financial damages, criminal penalties, or civil penalties?
Levin: Yes, and you would see changes. This would apply not only to board members but also to the CEO and possibly even the managers in charge of development. Why create more rules and bring in outside groups to inspect AI models? Just think about it. Anthropic is going public at a value of $2 trillion. If laws require good management and oversight, won't the company want to protect that value?
Kirsner: So you are saying that if Anthropic creates an AI agent that goes out of control, the company is responsible. If I create an AI agent that does bad things, I am responsible. But Anthropic should not automatically be held responsible for what I create?
Levin: Well, the question we need to answer is how many safety limits Anthropic gives users to prevent them from hacking or misusing its systems.
Kirsner: HubSpot is a sales and marketing technology company that has sometimes been the most valuable software company in Massachusetts. You advised the company in its early days.
AI has hurt HubSpot's stock price. In particular, companies are using AI platforms to do work that they once paid HubSpot to do. Last week, the company announced that its co-founder and former CEO Brian Halligan is coming back. He plans to turn the company into a major AI business. How would you explain the challenge? Are you confident he can put the company back on a path to growth? I do not want to say he has to save the company, but can he help it grow again?
Levin: Let me give you some background. HubSpot, Salesforce.com, and other customer relationship management platforms were part of the software-as-a-service, or SaaS, movement. I believe HubSpot did change after ChatGPT 3.5 arrived on November 30, 2022. But it did not change quickly enough.
Some customers still use HubSpot as a web app. But the company will increasingly divide its products into more traditional tools and AI-powered products. Farther in the future, it may become a company built mainly around AI, with Dharmesh Shah and Brian Halligan back in leadership roles. Shah and Halligan started the company together. Shah never left and is now its chief technology officer. I believe this pair has the energy, intelligence, and experience to move the company toward an AI-focused future during this second period of working together.
Kirsner: To be clear, do you own HubSpot stock? Are you buying more, or are you watching from the sidelines and cheering for the company?
Levin: I have owned HubSpot stock for a long time, and I still own it. I think it is an especially good investment now.
Kirsner: But you are not buying more because Brian is coming back. You are just keeping the shares you already own.
Levin: I am definitely keeping my shares because Brian is coming back.
Kirsner: Last week, people protested outside Google's Cambridge office because Google is connecting new AI tools and technology to Google Search, which schools use. They are saying schools cannot separate Google Search from Google's AI tools. What do you think will happen?
Levin: Google already includes AI-powered search alongside its regular search engine. I believe schools are simply not ready for AI in many ways. I teach regularly at Harvard Business School and Stanford Law School. One important point is that many teachers below the college level are not prepared to use AI. I believe they would welcome it more if they received better training. But right now, teachers are very busy. They have a lot to do and very few resources. So there are many reasons for their opposition. It may also be connected to the same feelings behind opposition to data centers. Many people simply do not want this technology around them. Finally, many people see AI as something that could cause job losses.
Kirsner: There is a lot of worry about this. It is not just among the students I have spoken with. Many people want to know, “How will this affect my job?” But I do not think the federal government, state governments, or technology companies are admitting that this fear exists.
Levin: The fear is very real. Microsoft offered 60,000 people the chance to take furloughs two weeks ago. Those workers are feeling it. They know that, to a large extent, their furloughs are connected to the arrival of AI.
Kirsner: Would it matter to the U.S. economy if Jensen Huang were no longer the CEO of NVIDIA?
Levin: I love the way you asked that question.
Kirsner: Trump seems to think it would matter.
Levin: Well, he is not an unbiased person. He has a $62 million investment in NVIDIA. NVIDIA is at the center of AI leadership, economic competition, national pride, and national security. That is a very complicated mix of issues.
It does matter who runs the company because national security concerns should lead to limits on chip exports. I believe these chips are an important national security resource. Competition with China, including economic competition and other issues, will help decide our future. We need a better administration that makes decisions based on knowledge and facts. It should not be filled with technology people who have their own financial interests. It should include people who understand world politics and diplomacy. That is what I hope to see in two years.
Kirsner: Should we try to control NVIDIA's most advanced chips in the same way we controlled uranium and plutonium during and after the Cold War?
Levin: Absolutely.
Look, Jensen Huang may have cheated, and he should face punishment for that. But there is another side to this story. There are many things we do not know, and people are rushing to judgment. A lot more information will come out. I also expect powerful groups, possibly including a Democratic-led Congress, to pass laws that deal with some of these problems. Those laws may also cover NVIDIA and data center rules. Perhaps all of this will make more sense when we have a much smarter Congress.
Kirsner: The last thing I want to ask about is local AI companies. There is Liquid AI, Suno, Blitzy, and a defense company called Shield AI. What do you think of the local AI scene? Are there other companies that you are especially excited about?
Levin: You are focusing on company names, but you should be focusing on the industry. Physical AI, especially robots, is our future. That is where growth will happen. We developed Roomba, the robot vacuum cleaner, years ago. Many years ago, we also developed Digital Equipment Corp., which worked in scientific computing.
Kirsner: By physical AI, do you mean bringing together our experience with robots and hardware with AI software and AI models?
Levin: Robots are basically AI systems with a lot of metal around them and the ability to move. Think about future robots that live with our grandparents and other older people. They could provide friendship, comfort, medical help, and many kinds of health checks. AI will power those robots. That is the future. The Boston area has everything it needs to bring together AI and robotics and become a major center for physical AI. This could be an enormous opportunity. All we need is support, sensible limits on regulation, room for new ideas, and encouragement.
