走进沙特的 AI 雄心：Tareq Amin 谈打造新的科技超级大国

I believe connectivity is a human right. I felt that this was a great opportunity to really build and enable a country like Saudi Arabia that has incredible potential and capability, to really build a digital champion and a digital hub for the region. This is the first time I've landed in a place where I feel I'm home.

I've been spending a lot more time in the Middle East, in the region. It's changing so dramatically year after year. Um, and I'm going to be I announced on the podcast 2 weeks ago, I'll be doing Founder University in Riyadh in November 3rd, 4th, and 5th. And I'm really excited to partner with such a dynamic country. Maybe you could fill us in on what's changed over the last 5 years in the region.

Well, thank you very much. First of all, I must say it's the first time I've done an event that's only 45 minutes away from my home. My flight from Riyadh, obviously, was 23 hours, but just to give you some context, I was born in Amman, Jordan, before coming to the US to finish my studies and progress in my life and my career.

I moved from Tokyo to Saudi Arabia, but this is the first time I ever worked in the region. Sometimes it depends on your perspective and how you look at the world. I wanted to really discover and understand everything I used to hear about: the massive opportunity for transformation and the diversification of the economy.

My first observation really started with the discovery of how amazing the people were in terms of hospitality and how welcoming they were. I really felt like I was back home. It felt very, very different to me. The second thing is that the population is awfully young and hungry for new things and new advancements.

It was a pleasant surprise because what I was really worried about, and what everybody keeps talking about, was that maybe Tareq has done great things in his life, but talent and opportunity may be a big challenge in his new venture. It was remarkable in every aspect, whether you look at it from the government side, in terms of society, or the transformation. It has changed dramatically.

It's really an area where I feel the opportunity is remarkable, and society's embrace of the future, in which digital and AI are fundamental to their transformation, is remarkable.

Tareq, can you maybe explain what Humain is and how it came to be? Because it was sort of an evolution, right?

For you to know, we take things for granted in the US to a certain extent. Let's assume you're a startup here and you want to access compute: it's as simple as going to a hyperscaler, and within 30 seconds you're ready to go.

The story is fascinating because my first discovery, in the role I was in at the time, was that I was hired to really run a subsidiary of Aramco—a new company that was intended to diversify out of its core business, in which digital and AI were really core components. One of the biggest surprises I had was the lack of AI infrastructure.

I did not know that the ability for startups and companies to access AI infrastructure was a challenge. Look how large Aramco is: it took them 9 months from the purchase-order process, export-control process, deployment, and installation. That means we hindered a company from really achieving its objectives by 9 months.

That was the first opportunity I saw that we needed to address in the country. I tell you the story because this is how the Humain story started. I started meeting with several ministers in the country. I told them, “I see nothing but opportunity. I see something that the country could really do and participate in on a global scale.”

I think we have an abundance of land, an abundance of power, amazing connectivity, and an opportunity to really participate in digital infrastructure. Last year, during my birthday, I got a call. It was a strange number, so I answered it. It said, “His Royal Highness wants to meet with you.” I said, “Okay, what did I do?”

Anyway, we all came in. I said, “What's the topic?” The topic was AI brainstorming. This was the first time I had the opportunity to meet His Royal Highness Mohammed bin Salman.

It was not about brainstorming. It was about really addressing what the country needs to do to accelerate and address some of the fragmentation that exists. The idea and concept of Humain is that, in order for us to really accelerate our development, we should bring public and private entities together, take AI investments, projects, and initiatives, and put them under one umbrella company that's really focused on the entire AI total value chain.

Honestly, in that meeting, it took about 10 minutes to come to the realization that the opportunity was massive, but we needed to unite the fragmented effort and put our energy into addressing the basics.

And you were given a pretty large balance sheet to do this with.

To start with, building infrastructure is not necessarily a small task. To do this, participate, and make an impact, you need to be very well funded. Even though we call Humain a startup, it's actually a very well-funded startup.

Between our data center segment, our models team, applications, and also our ventures that we will soon have in the United States, it's a very well-funded company. I'm feeling really comfortable about the opportunities that we could capture between capital, people, talent, and partnerships.

Should we think about this as—we live in a world, as you said, where when you're back home here in the US, there's AWS, GCP, and Azure. It's simple and straightforward, but the rest of the world is a little bit more complicated. Do you see Humain being that competitive alternative for the rest of the world? Is that how it starts?

I think, look at it this way: in certain areas, we partner really well, and all of them are colleagues and friends. I've known the CEOs of these companies really well through my tenure in Japan and India.

I will tell you, I think many companies do not see the Middle East the way I see it today, and they don't understand the requirements and the needs of what we have to do to build the depth and the talent. I see, in certain areas, great partnerships, and in certain other areas, I think that we will be a compelling alternative. But it doesn't mean that we don't partner.

Certainly, when you see the partnerships that we announced with the likes of AWS and Google, clearly I want to see the companies invest in addressing the lack of infrastructure that I saw. A portion of what we do today will be built with our own technology, and a portion will be done through partnerships with the hyperscalers.

Give us the lay of the land from the foundational-model side—the OpenAIs, the Anthropics, the Groks. How is that moving? How do you work with them? How do you make sure that everybody can get access to the latest models?

A couple of things we have done: for people in this audience who have participated in building a foundation model, it's not trivial and it's not easy. One of the acquisitions that happened at Humain is where I was really surprised to see the depth of the science team we have today at Humain.

The first comment I get on my social posts is, “Yes, you have capital, but there's no people, no talent.” So we ended up building a foundation model, and I'll tell you why we did this. We built it from scratch. It was not distilled from any open-source tool.

We did it for 2 reasons. First, let's venture into this to understand the depth and capability of the organization. Second, when it comes to culture, language, and biases, I felt that it was important for us to really participate in this area.

So we launched in Saudi Arabia what we call Humain Chat. I just wanted to see the reaction from the people who use this model. The only objective of it was an Arabic-first preference: that's the way we train this model, not with an English-first preference.

On that, you see huge differences in how the foundational models process and digest data?

Yes. For us, in any model that's used today, the quality of data is always much better than just feeding it quantity of data. Our model today has a proprietary set of data in the Arabic language that you cannot find on the public internet.

We had a proprietary data set that was important. Obviously, within the Arabic language, the government today uses all its correspondence, transactions, and translations in Arabic. So the preference for training first in Arabic is very different from starting in English and then adding the Arabic components.

We built that and launched it. It actually became the number-one app in the App Store in the country. It's a really important thing for us. I'm going to be really specific: we didn't do this to say we're better than OpenAI, or we're better than X, or better than anyone else. It was important for us to train the team on understanding how to build the entire stack.

It's really critical. I would tell you there's nothing else I would do differently than what I've done over the last 9 months in building this foundation model.

We have great partnerships today, with discussions with OpenAI. We use Anthropic a lot for our AI coding tool. So, I see this is not an option of using this or that. I think we are trying to formulate a strategy around the model, but I will tell you where we differentiate—hugely differentiate. I think we have found the answer to what I heard everybody talking about.

I am a huge believer that AI is not a bubble. The reason I believe this is that I think we are one of the few companies that have found true value realization, and this is a really mind-blowing story.

When I took on the CEO role of Humain, I had an option to say, “How do you run a large enterprise? Do you run it the same old way, where you have legacy systems in play and hundreds of IT tools that do the job for finance, legal, HR, and cybersecurity, or do you do something different?”

What we have differentiated on, while I'm excited about the model, is what we build on top of the model. We will launch in October this year a platform we call Humain One. It is truly the AI operating system for the enterprise.

Imagine the era—I believe Windows was invented in 1981. We all got taught that you use icons on the desktop. If you want to take a vacation, you go to SuccessFactors or another HR tool. If you want finance, you go to Oracle or SAP. If you want something else, there's another tool, another application, another icon.

Now we change this enterprise world into a really intent-driven system, a multi-agent orchestration system, and its impact is unreal. What has happened in driving true value realization, at least in my company, I could tell you is remarkable. The efficiencies that we have derived from changing these legacy systems are incredible.

Maybe the biggest challenge that I had had nothing to do with the technology—nothing. This is perhaps my own personal opinion about the struggles that companies are going to go through as they embrace AI into their operations. It's the mindset, mentality, culture, and organization. It's an underestimated effort.

So, between the model and what we build, I am more excited about what we have done on top of the model.

Let me ask you a pointed question. You're in Saudi. There's an enormous amount of energy, almost a surfeit of energy. Do you have this unfair advantage where, if you need a vast amount of energy to throw at a compute problem, you effectively get it for free, or do you have to pay for it as well?

No, of course not. Even though it might seem like it, I wish I could get it for free. That would change my business model dramatically.

When I took this role, I think one of the key things I said was, “What can we really compete on, and what can we do to offer the world something that the world desperately needs today?”

I think everybody in this room realizes we are really just at the beginning of what I would call the AI inferencing world. Model training will continue to evolve, but as we move into accelerated compute, I think the demand is going to become much larger than it is today.

So what's the challenge? The challenge is really simple: power. How do you really find power? That's where I think Saudi Arabia has a big role to play. I was bullish enough to tell the world, “Outside of the United States, outside of China, I really think Saudi Arabia has a good shot at being the third-largest country in AI infrastructure.”

We have to go through processes to secure power from the Ministry of Energy through the local electric company. I am treated as fairly as any other entity that comes to the country. My rates, my tariff, are equivalent to what Google would get and what AWS gets.

However, the energy generation that exists in Saudi Arabia is just remarkable. Maybe Saudi Arabia today leads the world in energy exports via oil; we should look at an opportunity to lead the world through energy exports via tokens. That's really a key area where I think we could differentiate: energy and infrastructure.

Let's talk a little bit about the talent pool there. This was shocking to me when I got there, but so many of the young people, even up to Gen X—our generation—have been educated in the West. In fact, the woman with whom we are in partnership at Sanabel has an MBA from Stanford and Oxford and all these incredible schools.

They've all come back after these incredible scholarships that are given by the country. Talk about the talent that has come back to Saudi, the application level, and golden visas as well. Just how open they are to having companies set up shop there.

And there's a bit of competition, is there or not, between Doha, Dubai, Abu Dhabi, Riyadh, and Bahrain? There's a really great competition emerging for talent. You've got the entire country, all the nationals, very engaged in starting companies.

This is such an important point because I think this is a key fact that people don't realize yet.

I didn't, yeah.

Yeah, if you look at 30 years ago, the investment and the strategy for Saudi Arabia to send many of its citizens abroad, across the world, to pursue degrees is really paying off in a really large way.

Again, going back to the example I mentioned about my models team, I have 40 PhD scientists today on my team. Sometimes, honestly, when I sit with them, I say, “Maybe I don't have enough credentials to sit with you guys in the room.” They graduated with PhDs from Stanford, MIT, Harvard, Oxford—from anywhere that you want. They've really obtained the foundation knowledge to take that next step.

When they graduated, were they thinking about staying in the US, getting an OPT thing, or was there this pull to go back and say, “Okay, we're now trained. Let's go back and help our mother country”? What is the motivation of these individuals?

A couple of things happened. This is my own view, and I talked to my colleagues in Saudi Arabia. I really think that they had a sense of responsibility to take part in the building process of what tomorrow is going to look like. I would have done the same if I were in their shoes. I would have gone back and really helped with this build process.

They learned a lot, mostly from the United States and the UK. Now they all came back looking at the opportunity because the inflection point has happened. If you look at the fast-paced transformation, it's an opportunity I don't think they want to miss. It's really changing rapidly.

Part of that maybe is a desire to work in the era of MBS. Maybe you can use that as a jumping-off point to tell us: What is it like to work with this person who's young and beloved?

A couple of things. I'll tell you my honest opinion. First of all, the first meeting that we had with him—my key observation and takeaway was that his vision blew my mind. I don't think I have seen someone with such a bold vision of what the future looks like.

His sense of responsibility to the citizens of Saudi Arabia, and the diversification and transformation initiatives that are happening, are second to none. It's really incredible.

I was delayed yesterday. I was supposed to be here, but the man doesn't sleep, really. He's on all the time, thinking about initiatives that need to be done and fast-tracking them.

Everybody now is really rallied. If you go into Saudi Arabia today, one thing that is surprising is that when you take a taxi cab, go to a hotel, or talk to a government official, everybody is intertwined with the Vision 2030 mission. Everybody.

This is something that rallied the society around an idea, and the idea has incredible merits.

And is it both the old and the young?

It's—I mean, what I'm seeing now across the board, even when I talk to the older generation or the younger generation, this mentality shift and transformation has happened.

I'm not representing everybody. I'm telling you about the segments I'm interacting with. I see nothing but optimism. But you guys know this much better than me.

Such a key point. That was what I took away when I started spending time there over 3 years ago. I saw the change every year. The enthusiasm level reminded me of New York and San Francisco in the '90s and early 2000s.

The opportunity to build anything—the culture was on fire. You've got movies and cinemas coming back and all this great stuff happening. They're so engaged, and they want to learn how to build businesses.

They've got this real sense that Vision 2030 has inspired everybody to think, “What can happen? What's possible?” We have this incredible energy business, sure, but as they explained it to me, we can take that energy business and, over the next 5, 10, or 20 years, transform this entire region and be a global leader. It's really inspiring.

And to add to your point, going back to your question about competition, this is really interesting. There's a lot I have learned, and I owe a lot to what I learned in the US—quite a bit, a lot.

I learned innovation here. When I moved to India, I learned scale. When I moved to Japan, I learned precision and quality. You come to Saudi Arabia, and I'm learning optimism and vision, truly.

Hopeful. You know, I find it to be really remarkable. I'm living there almost full-time.

What I have done, if you come to the HUMAIN office today, in my opinion, it is no different from a Silicon Valley office.

Not at all, yeah.

It is. So, I encourage you: Come visit and see how HUMAIN is built from the basics of the basics—how the building layout is, how the open-office design is—to remove barriers and mentalities. And I want my team to become product creators, not just resellers. So, that’s very important.

Tareq, as you are building your infrastructure, there are supply chains and strategic partnerships and relationships that I’m sure emerge here in the United States, but also in China. Saudi Arabia seems to be in this really interesting position as a large energy supplier, as a large partner in capital, and now in building infrastructure that could create tension amid the global rivalry between the US and China. How do you think about managing each of those two markets, establishing relationships, and deciding where you align yourself?

So, if you see how HUMAIN was launched—and I wish I could take a small amount of credit for this—but I asked for one thing: Before President Trump came to visit Saudi Arabia, we should launch the company at 3:00 p.m. Because we wanted to ensure that the alignment we are going after is really where the innovation is, where the talent exists, and where the infrastructure exists.

Let’s be really realistic and truthful: Today, the US is leading, especially on the semiconductor side, and we don’t want to miss this opportunity. If you look at the partnerships we have done with AMD, with a startup called Groq, with NVIDIA, and with Qualcomm, it shows you clearly our commitment to this relationship and partnership. It’s really deep—not only on the silicon side, but the same thing you’re going to hear very soon on the AI and software-application side.

That’s why HUMAIN, to succeed, really needs deeper engagement and relationships. I spend a lot of time with startups in the US. To be very honest with you, the hope and optimism I have—and David and the team have been doing a really great job with this—is our ability to say, “We understand very, very well the concerns that one might have.”

But if you look at the partners that we have selected, from the software layer in the cloud to manage tenant management, to security in the data center, to making sure that these servers are secured, we will do everything that is required, with the optimism that HUMAIN will be thought of as a trusted supplier for the US.

David Sacks, how important is the American and the kingdom's relationship globally and for humanity in terms of the president's agenda and his prioritization?

Well, it’s been a critical relationship for the US and for Saudi Arabia since, I think, 1945, if not before, when the founder of Saudi Arabia, King Ibn Saud, met with our president—our king, so to speak, FDR—on a battleship, and they hammered out the foundation of the modern world: The US would provide security for the region in exchange for the steady flow of crude.

I think FDR did that on his way back from Yalta. People don’t know as much about that as they do the Yalta meeting, but that was a very important understanding. The relationship evolved over the last 80 years.

I went with the president on the trip to the Middle East in May, and, number 1, like you said, the business culture in Saudi Arabia is very Americanized. Many of the Saudi elite have studied in the US. Second, they want to have a good relationship and a partnership with the United States. There’s nothing competitive at all about that relationship.

Third, when it comes to high tech and AI, they want to be part of our technology ecosystem. When I got back to Washington, I was really surprised at how controversial it was that we wanted to do business with the Gulf States, and Saudi Arabia in particular.

The way I see it is very simple. Saudi Arabia is going to have data centers, of course. Every sovereign country that can afford them is going to have data centers. Is that going to be American technology or Chinese technology? It’s basically going to be our companies, or it’s going to be Huawei.

It’s binary. And why would we want to push any country into the arms of Huawei?

Especially when their preference is to work with America. Well, I think it’s especially for 2 reasons. One is the proximity that that region has to 4 billion people.

Because if you think about building data centers, ultimately, for inferencing, there’s a certain window of time, and you have to be under several hundred milliseconds. But if you do that and draw that radius, you’re counting half the world’s population. We can’t strategically ignore that. Otherwise, you’ll be forced to do it with somebody else.

We’ll get to Tareq in a second, but I think these restrictions on the region were placed in October 2023 by the previous administration. At the time, it was justified on the grounds that the US was the only game in town. We were the only ones who could really make advanced semiconductors, so therefore we could impose whatever restrictions we wanted. Nobody would have a choice.

But since then, if you’re reading the headlines over the past few months, it’s all been about Huawei, Cambricon, and SMIC. China is rapidly advancing. Dylan Patel from SemiAnalysis just had a report that by next year, China is going to be making millions of chips. Admittedly, they’re not as good as American chips.

But if we deny the rest of the world the ability to participate in the American tech stack, then they will participate in the Chinese tech stack. I think the question of what we sell to China will always be a complicated question for obvious reasons. But when it comes to the Middle East and the rest of the world, I think it should be an easy question: As long as these countries are abiding by our security requirements and they want to be partners and allies of the United States, we should allow them into the American tech ecosystem. Otherwise, we’re just creating a Huawei Belt and Road.

Maybe one important thing: We got approval last year, and this is a really important case study for you. When we looked at our business model, the number we used was exactly 4.4 billion.

4.4, sure.

We could reach 4.4 billion people within the tolerance of fiber latency, so inferencing would still be really ultra-responsive. We got approval after a few months for a startup in the US. This is a classic example, and you could really validate what I’m going to tell you now. We supported this US startup—look at where they are today and what their valuation is post-engagement with us.

At that time, we picked Groq for inferencing because I thought it was a really interesting startup that could help us democratize the cost of inferencing. At the data center that we have deployed at this end, we have about 19,000 of their chips deployed. 130 countries are now using this inferencing cluster. Five percent of the traffic is from Saudi Arabia; the rest is from abroad.

Why? Because we’re able to offer differentiated inferencing costs, and we offer that to the world. Keep in mind what we also did for security guarantees: We said, “GroqCloud, you manage this.” So now I don’t have to worry about KYC requirements. You follow US rules. All I care about is that I have revenue that I participate in, and it was really a win-win for both of us.

I hope we’ll repeat the same thing with the deployment that we’re doing with NVIDIA and AMD, and then do something really exciting with Qualcomm on the edge as well.