Marc Benioff, the co-founder and chief executive of $262bn software giant Salesforce, always has a lot to say.
The decabillionaire is Silicon Valley’s arch salesman and a ubiquitous figure in San Francisco. His 61-storey tower dominates the skyline and Dreamforce annual conference takes over the city for a week in September, where he is flanked by celebrities like Will.i.am and Matthew McConaughey.
Benioff’s fortune has allowed him to become a serial angel investor and mentor to dozens of entrepreneurs in tech. He worked under Larry Ellison at Oracle for a decade before setting up the cloud software group in 1999. In 2018, he and his wife Lynne bought Time magazine from Meredith Corporation for $190mn in cash.
Most recently, he has bet the future of Salesforce on artificial intelligence, counting on digital agents to revolutionise everything from customer service to supply chain management.
In this chat with Stephen Morris, the Financial Times’ San Francisco bureau chief, Benioff discusses how AI is disrupting work and every aspect of our lives. The audio of the conversation is included.
Stephen Morris: Marc, you have been around Silicon Valley and the Bay Area for a very long time. You were born here, you run one of San Francisco’s most visible companies, and have a better perspective of what’s going on right now in technology than almost anyone else. So how can you characterise this moment, with AI looking set to disrupt almost every industry around the world. It’s an exciting but overwhelming moment.
Marc Benioff: Well, you’re right. Our industry hits these moments. Every now and then, there are these Steve Jobs moments where a whole new part of the industry gets invented. So, it could be Tom Watson with the mainframe, it could be Steve Jobs with the personal computer, it could be Marc Andreessen with the browser. We’re in a moment where we see a level of innovation and capability and funding, and all the magic of Silicon Valley.
You’re seeing it first-hand. You must be thrilled to be in Silicon Valley because you can’t understand it unless you’re right there seeing what’s happening. And that magic is happening mostly in San Francisco, which is kind of cool. It is a big, big deal.
SM: So, the flip side of that is, all the money, hype [and] promises: you’ve been through bubbles before; is there any part of you that worries that AI is just the latest iteration of this?
MB: If you don’t have all those components, you’re not going to get the breakthrough. Because you’re going to have to throw a lot against the wall before you figure out what sticks. And that is the magic. So you’re like, “wait a minute, this doesn’t make any sense. These 10,000 companies just got funded, and only 100 of them are good.” But you have to fund them all to get to the 100.
We saw that in the internet revolution — we didn’t know who was going to be Google and who was going to be Ask Jeeves and Yahoo. It’s a progression. Tactics are going to dictate strategy over time. But you’re in a moment now where there is a huge cycle. If you’re not getting a hype cycle or funding cycle — or you could call it a bubble, you could call it a lot of things — you’re not going to end up with the incredible outcome on the other side.
SM: You don’t just run Salesforce, you also run Time magazine. Recently, you put DeepMind founder Demis [Hassabis] on the cover of your 100 most influential people. He is one of the, if not the, foremost scientist and proponent of technology, whether it’s through AlphaFold or the AI Lab DeepMind. He has some pretty stark warnings about what could happen if humans lose control of this technology, or bad actors deploy it maliciously or unwisely. Where do you sit on this ‘Doomer versus Boomer’ AI scale?
MB: Well, I don’t make the editorial decisions on Time magazine, but I do own it. And you’re right, we made Demis one of our 100 most influential people in the world. That might have also been influenced by his Nobel Prize. People like Demis, especially those in the core AI world — it could be [AI pioneer] Geoffrey Hinton, it could be so many others: I have a list of people who are, a) incredibly innovative, very excited, and then, b) also, hey, there could be a dark side to this technology.
Technology itself is never good or bad. It’s what you do with it that matters. All technology can be used for good and bad. And that could be genetic, nuclear or AI technology.
We could list the movies that go through the good and the bad scenarios. But it’s good to have our eye on all these things. One of the people that I work closely with is Peter Schwartz, who’s our futurist, and he was a key contributor, writer on Minority Report, on WarGames, on Deep Impact.
When you look at movies like those about the future, yes, there are dark scenarios. That’s how we try to figure out how to get to the right place. We have to talk about what could go wrong also.
SM: Are you more optimistic than pessimistic?
MB: I am naturally more optimistic. But I think that people do look at what these potential dark side scenarios are and make adjustments. Even our friend Elon Musk, who is kind of an amazing visionary, he started a couple of years ago with an incredibly dark vision of AI. He’s become more optimistic over time.
So things are constantly changing. It’s how people are using it and how they’re adjusting. Look, none of us want a Hiroshima moment. Nobody wants a moment where the technology is used in a horrible way, and people end up dying. That is the worst possible situation. Or a human-engineered virus that ends up killing a lot of people. Nobody wants these things, and nobody wants a dark AI scenario either, where you lose autonomous weapon systems.
People want to be happy and healthy and take care of their children and their families. I know you do, and I do. And that’s what we want. We want to make sure that technology is used for good. But there are a lot of different actors in the world, you know that.
SM: You have made a big bet with your own company on AI. We’re not talking about killer robots or genetically engineered viruses here, but you, last year, relaunched it under this brand Agentforce. I’m wondering how that is going. There’s a lot of talk about the potential of the technology and a lot of questions about the revenue generated on all sides of the equation, from the model builders to those deploying agents. Are you seeing meaningful customer pick-up and a tangible increase in your revenues as a result?
MB: We are. We’re building a multibillion-dollar business on this simple idea, which is we are seeing what is going to be a multitrillion-dollar opportunity in digital labour. That’s unfolding before our very eyes. This idea that the world will end up spending $3tn to $12tn on forms of technology, including agents like you mentioned but also robots and others, to provide digital labour, to usher in this new world of abundance: that’s a very real idea.
SM: Could it also be a threat to your business model? For a long time we’re really locked in to big companies through [customer relationship management] systems selling software as a service. You helped companies organise their data and provided them with software solutions. We’re getting to the point now where you can ask AI to design a piece of software or piece of code. You can give it access to your company’s data, direct to one of these model start-ups, sort of obviating the need to buy these big systems, which for years have underpinned your business. When you started Salesforce in 1999, you said it was the end of software. How are you thinking about this moment? Could it be the end of SaaS [software as a service]?
MB: It’s always the end of something and the beginning of something new. But you hit on it when you said these agents are going to have access to all your company’s data. And for our customers, banks, insurance companies, governments, tech companies, media companies, they’re all managing their data. And that data is a federated resource. Federated means connected, a resource inside their company where they’ve connected these data sources together. Because inside a big company, you don’t have success if that AI isn’t fed with the right data.
SM: One of the issues that a lot of people are concerned about on the more social side is that certainly there’s a lot of productivity, and potentially a lot of money to be made, through selling companies these agents. But there are going to be some job losses over the next decade as these models and agents become more capable. Do you think government and society is doing enough to prepare for this upheaval?
MB: I don’t know if it’s an upheaval. I would say that it is important to realise that, when new technologies appear, there can be changes in the workforce. And this one is no different, especially when we talk about the emergence of digital labour: you could have fundamental productivity increases in GDP without adding more people. That hasn’t happened too many times in history, so that’s very exciting.
That can get to a new level where an entire country may be more successful and more productive because they’ve deployed the right level of capability. You talk about the US going through this transformation, wanting to bring back more manufacturing than ever. It’s going to have to come back through digital labour because we don’t have people to put on those factories. So it’s going to have to be a robotic and an agentic layer to make those factories more successful and deliver those products.
You’ve been in a Waymo and so have I. There was no one driving. I’m sure the first time you were in there, you’re going, “wait a minute, what’s happening?”
You can call the car, it comes and picks you up. There is no one in the Uber, it’s called a Waymo, and off it goes. It’s a robot on wheels. That’s digital labour. This idea that our society is going to change with AI and transformation, that’s happening, and we have to be ready for it. We have to think about what those implications are. And we shouldn’t try to hide those conversations. We should say, yes — just like it might have gone from the horse and buggy to the car, to the taxi driver, to the Uber driver, to the Waymo.
SM: You touched on the biggest topic in the world: tariffs and the economic policy of the second Trump administration. US tech companies, including yours, have enjoyed a de facto global dominance for decades now, based on the ideas of free markets and globalisation.
Is there a danger to incumbents, such as yourself, that the shock and awe of these tariffs, the policy flip-flops, the attacks on the entire principle of globalisation, the singling out of China, will lead to a general loss of confidence of the rest of the world in American decision-making and stability?
MB: Every single one of our customers has to wrestle with this new reality, that we have a democratic system in the US so we can get a new president. And that president can come in with different policies and ideas, and that those presidents change over time.
So our customers have to be ready to change. Rarely does a day go by where I’m not helping customers manage their tariff situation. It requires a huge amount of data management. Guess how they’re going to manage that. They’re going to manage it with our technology. That’s an extremely important part of what’s happening.
It’s a shift in where we were and where we’re going. But those shifts are happening all over the world all the time. That’s what makes life more interesting. If everything stayed static, it would be quite boring, and I guess there wouldn’t really be a use for or need for the FT or Time magazine.
SM: Salesforce hasn’t been immune from the market sell-off, and its stock is down 25 per cent this year. Is this just market moves, or is this also a reflection of fears that customers, of every company, both US and international, are starting to get nervous about spending in this economic environment, and they might cut back on services?
MB: We’re going to rise and fall with the market, there’s no question about that. We delivered a quarter with record revenues and profits, and guidance that was appropriately conservative. Our cash flow was as strong as it’s ever been, I think one of the strongest cash flows of any company in the history of business.
We have to continue to do what we are doing, which is making our customers successful. Nothing is more important than that, and guiding them through this incredible period of change and transformation. They’re going to build all kinds of new systems and capabilities.
SM: Could you talk more broadly about the changes in politics? Many, but obviously not all, of your peers in Silicon Valley campaigned for Trump from the venture capital community. We’ve seen lots of tech CEOs like [Meta’s] Mark Zuckerberg and [Google’s] Sundar Pichai try and rebuild bridges with the new president. You’re a Democrat, you’re a big donor. How do you explain the changing national political environment, but also the relationship of big tech with it?
MB: Well, I’m not a Democrat, but I do live in San Francisco. I’ve never been a Democrat, which might surprise you . . . I worked in the Bush administration.
I was the chairman of the president’s IT Advisory Committee. That was my round of work with the presidential administration. I was glad to have my time in the barrel with that, and I’m grateful for others who are willing to do their time now.
In California, it’s heavily Democratic, and in San Francisco. But some of that could change for a lot of the reasons that you’re talking about it. Things change, the world changes, political tastes do change. Things are becoming more conservative. People want certain issues addressed that were not being addressed by the Democratic party, which is why you saw such a significant red wave.
After my time with the Bush administration, I became more independent and decided that I wasn’t going to become a Republican or a Democrat, that I was going to be more of an American. When I bought Time Magazine, it reinforced that for me, because I decided I wasn’t going to make political contributions any more.
It’s been a transformation for me. When I think back, it’s been about 20 years since I was working in Washington and working on things like cyber security, healthcare and computational sciences. When I look forward now, over those two decades, it’s a completely different Republican party.
I don’t think I’ve really resonated with either party for quite some time. I’m happy to be more in the middle.
SM: How were you navigating this? Have you been to Mar-a-Lago or the White House yet?
MB: I haven’t done that because I do need to keep that political balance. I’ve been cautious on how I’ve approached not just this administration but also the last several. Because since I bought Time magazine, I thought I needed to be more impartial.
SM: We know journalism is an increasingly difficult job. You’re the owner, not the editor, but you still have to ensure that the publication you own walks this tightrope between robust, honest and potentially critical coverage, while also not coming under attack from the Trump administration or anti-mainstream media figures like Elon Musk. That’s a difficult path.
MB: Oh, I have a good story for that. Time magazine has been around for more than 100 years. We do a Person of the Year. We have another one called our Time 100. Person of the Year was Donald Trump this year. We have a tradition of putting the elected president on the cover and saying he’s the Person of the Year, because it’s a strong case that he is.
I can’t tell you how many angry emails I got from friends saying that I should have never done such a thing. And, of course, they’re very big, heavy Democrats. When we did that Person of the Year cover, we sent one of our best photographers to Mar-a-Lago, our editor, our CEO, a whole team. We made sure it was an excellent cover, maybe one of our best ever.
If you go to Mar-a-Lago, which I haven’t done: friends of mine have sent photographs of walls of Time magazine covers. I think Donald Trump will soon be on Time’s cover more than any other human in history.
That idea that it doesn’t matter what we write — and I’m sure you have this experience yourself — you’re accused of being too left or too right.
I just ignore it. Our editors do a remarkably good job. We want to be objective. We want to be more centrist. I’m sure that because some of them are in New York City, they are more left-leaning than they realise. But it’s hard to be a journalist today and not be aligned with one political party or the next. And yet, that has to be a key goal of being part of a major media organisation.
SM: One of your better qualities, from a journalistic perspective, is your outspoken nature. And there are a few issues you haven’t sat on the fence on. You’ve been critical of some of your rivals’ products, in particular Microsoft Copilot, which is their version of an agent, based on OpenAI’s GPT models. You flat out called it Clippy 2.0, which ruffled quite a few feathers in Redmond, with one of their executives firing back: “Marc doesn’t know what he’s talking about.”
You talk frequently about the breakdown in the OpenAI and Microsoft Data Centre relationship. Tell me what’s going on here. What’s your beef with Satya [Nadella]?
MB: I noticed that those executives stopped saying those things when it turned out that what I was saying was right. There are a couple of things you’re hitting on, probably two of the most important. Number one, you’re right, Microsoft did disappoint a lot of customers with Copilot. It looked more like Clippy, customers didn’t find value in it.
In the developer productivity area we saw this incredible new company Cursor appear out of nowhere. In the middle of Silicon Valley, you saw OpenAI announce they’re going to buy Cursor’s competitor, Windsurf, for $3bn. This is because Microsoft’s Copilot and GitHub also failed, so it didn’t deliver the level of productivity that it could have.
The second point which you hit on was this idea that Microsoft is just a ChatGPT reseller. That’s their AI strategy at its very core. They’ve become very frustrated with that. So they hired Mustafa Suleyman, Demis’s partner in DeepMind, to run a new Microsoft AI division, to build a new model, which is part of their Prometheus programme. This idea that they’re going to have their own models, and not going to have ChatGPT at the heart of Copilot.
And you saw it also play out where Sarah Friar, who’s a former Salesforce executive, now the [chief financial officer] of OpenAI, did a presentation at a Goldman Sachs conference. She laid out a stack diagram, which is the traditional tech way of explaining your strategy. In the data centre level and the application level and the API level and the model level, there wasn’t any Microsoft software at all. And the Azure data centres weren’t even mentioned.
I noticed that Microsoft’s executives [have] stopped saying I don’t know what I was talking about. It’s all playing out right in front of our eyes that there is a huge breach between Microsoft and OpenAI. It’s a full proximal rupture. And it’s not coming back together.
SM: Within that, there’s a larger criticism. You’re saying that these large language models, whether they’re built by Anthropic or Google or OpenAI, are becoming increasingly commoditised? Customers ultimately don’t care which model they’re using. They consistently leap over each other in capabilities and league tables. OpenAI, in particular, has a $300bn valuation, and some of the others are not far behind. Then you have the astronomical sums being spent on data centres, power and the other infrastructure needed. Are you worried about a bubble in that particular side of the tech sector? And if it pops, what’s the collateral damage?
MB: This is one of the most misunderstood aspects of this entire AI revolution, which is the role of the model. AI has got to where it is over the last two decades through open source. Salesforce is a huge contributor to the body of work, including the prompt engineering and other critical parts of these model technologies. Open source has made all of this possible. It is the driver of the innovation.
To that point, let me add that DeepSeek is probably the most fundamental transformation this year. This incredible [Chinese] model came along, an open source model with an MIT licence, which means it’s basically free, and you can put it in your product. And Salesforce, for example, could easily do this and reduce our cost of using our model by 90 per cent because they came up with very innovative new ways to deploy models that will save companies billions.
The current commercialised models had not come up with these approaches. They were moving to these. It’s a transformation from a technical model called transformer to another technical model called MOE [mixture of experts] that’s like, “wow, if we do this, we’re going to save a lot of money”.
It’s forcing other companies — not just Meta but all of them — to have their own open source model. A lot of this magic and capability is available for free in open source. And so you’re right when you say, “hey, don’t you think that these companies should be worried about this or that piece?” But at one level, they offer an open source-capable platform that’s maybe interchangeable. At another, they offer a consumer service that’s branded.
I use ChatGPT, Google’s Gemini, You.com, which I have an investment in. There’s another one called Anthropic that Salesforce has an investment in, their product is called Claude. And then we make some of them available inside our product Slack [which Salesforce acquired from Microsoft in 2021]. But you don’t even know which one we have there. We just pick the one that’s best for us and our customers.
The model will increasingly be a commodity. They’re all mostly doing the same thing. I’m sure an impartial observer would agree with that. And they’re following each other within six months. The companies are great. I’m thrilled they’re doing the work, and it’s an accelerant on AI. But the ultimate accelerant is that the vast majority of the work has been in open source.
SM: Now, your first job after graduating was for Larry Ellison, Oracle. He became sort of a mentor to you over a couple of decades. He’s made a big bet on Sam Altman and OpenAI alongside [Masayoshi] Son of SoftBank. Are you in contact? Do you still chat about AI and business? And what do you make of this Stargate partnership to spend half a trillion dollars on data centres in the next three to four years?
MB: A huge mentor for me, Larry Ellison. Incredible person. I started my own software company in high school. Then, I worked at Apple. I wrote some of the first assembly language natively on the Macintosh in 1984. And then in 1986, I went to work for Oracle for 13 years. He had a huge impact on my life and taught me all kinds of amazing things that I still use.
We talked about Mar-a-Lago in Florida. He lives right there, in the heart of everything that’s going on. It’s quite amazing. I’m just stunned. He’s 80 years old, and he’s as innovative and as pioneering as I’ve ever seen. And you’re right, he just announced this new Stargate data centre, where he’s going to help Sam Altman separate from Microsoft, and offer him the technology platform and the data centre platform.
That idea that he found this company, Crusoe, who was building this data centre in Abilene, Texas, and then helped, with his friend Masa Son who runs SoftBank in Japan, to put together the $500bn in a partnership with OpenAI. No one saw that coming. And it was a White House announcement. It was a stunning thing. Only Larry Ellison could have pulled it off. I am as big a fan and admirer of his as ever.
SM: Do you think there’s an inherent risk? You were early in saying there’s a lot of money and power going into these data centres. But we also see analyst reports every week saying Microsoft, Amazon, Google are pulling back on this and that data centre. There does seem to be a softening and real concern about overspending.
MB: We talked about this in Davos. It’s all played out, just like we thought it was going to, which is the models are commodities. They’re going to be relegated to certain consumer services. Enterprises are going to have their pick of the litter in terms of choosing what is best for them. It’s going to get dropped down into this line of business solution.
Our job at Salesforce is to make those solutions, not just help them pick the right technology but to future-proof them. They’re choosing our products and platform so they don’t have to make that model decision. They’re going to be able to choose what model they want over time and change it if they run it on the Salesforce platform.
That’s the heart of how Agentforce is built. I’ve put my philosophy into my technology as well, so that my customers won’t get burned as winners and losers are chosen by the market over time.
This transcript has been edited for brevity and clarity