10万亿美元 AI 基础设施建设遭遇光子学瓶颈
Molly O'SheaHerwig Van HoveYannick De Koninck
AI 的下一个瓶颈是连接 GPU 的网络，而不只是 GPU 本身。 模型已无法装进单个 GPU，智能体系统还会在服务器之间发起对延迟敏感的调用；Yannick De Koninck 表示，这要求网络“更扁平”、连接更密集。他最有力的判断是：AI 工厂约一半由 GPU 构成，另一半则是连接这些 GPU 的技术。
光子学之所以变得不可或缺，是因为就在 AI 工厂需要大幅增加内部流量之际，铜互联正接近极限。 光互联能够提供高得多的数据传输带宽；Herwig Van Hove 则将数据中心能耗定义为这场酝酿数十年的光子学范式转变的“杀手级应用”。
可投资的稀缺性横跨整个制造链条，而不只在光子学本身。 TSMC、GlobalFoundries 和 Tower 已投资硅光子晶圆，但“如果你把一片晶圆交给 NVIDIA，那并不是它订购的产品”——NVIDIA 要的是最终芯片，而最终芯片仍需要激光器、设备和衬底。吞吐量和良率仍然不足，原材料供应集中在少数供应商手中，东西方供应链割裂又进一步放大了风险。
NVIDIA 的加码把一场预期已久的转型变成了眼下的产能问题。 Van Hove 将需求冲击追溯到 ChatGPT 于2022年11月爆红，随后 NVIDIA 在2024年和2025年释放更强信号，并于2026年投资 Lumentum 和 Coherent。他称 NVIDIA 在当年前6个月投入的资金相当于整个市场规模的4倍，这揭示了供需错配：“创新已经就绪”，但供应链还没有准备好。
Themaa 正利用收购的一家欧洲晶圆厂，将新设施建设压缩到2027年投产、2028年全面爬坡。 该工厂距 imec 20公里，既有现成基础设施，也具备光子学和工业人才。客户要求的产量已经超过 Themaa 所谓“极其激进的预测”，但设备可得性、机器性能和衬底仍是约束。
欧洲的光子学机会确实可信，但真正的考验在执行文化。 比利时拥有异常深厚的研究密度——Van Hove 称，2000年至2015年，根特大学贡献了全球4,000篇光子学同行评审白皮书的35%——但欧洲并未持续地在本地将自己的创新商业化。Themaa 希望将这一技术基础与快速试错、果断决策的运营文化结合起来，部分参照 De Koninck 所描述的 NVIDIA 工作方式。
1. AI 工厂将互联升级为核心算力基础设施
Molly O’Shea 以 Tony Kim 对10万亿美元基础设施重建的估算切入：传统数据中心耗电耗水过高，也不是为当下参数量庞大的 AI 工作负载设计的。嘉宾随后将问题收束到这些设施内部快速增长的流量。
De Koninck 的逻辑很直接：模型“已经无法装进单个 GPU”，因此计算取决于 GPU 之间的通信。随着处理器以更高速率要求处理更多数据，铜互联跟不上，行业正转向带宽高得多的光连接。
智能体 AI 加剧了这一转变，因为模型可能调用运行在另一台服务器上的工具，并等待返回结果。这让计算之间的延迟变得关键，也要求行业用“更扁平的网络”替代深层级的层次化交换机树：GPU 与交换机之间需要更多直接连接，互联网络密度也必须远高于传统超大规模数据中心。
他的收尾判断让市场边界变得清晰：“其中一半是 GPU，另一半则是连接这些 GPU 的技术。”因此，光学与光子学解决的是与计算引擎本身重要性相当的基础设施问题。
2. 功耗是光子学的杀手级应用，但制造才是瓶颈
数据中心 CTO 被问到最让其夜不能寐的问题时，第一答案就是能耗。CEO 要求继续增长，但电网建设无法按所需时间表到位；Van Hove 举例称，Google 与阿姆斯特丹谈了3年后，最终把计划转移到了比利时。
转向光子学的必要性在2022年11月变得显而易见，当时 ChatGPT 迅速走红。NVIDIA 随后在2024年和2025年逐步加大光子学布局，并通过2026年对 Lumentum 和 Coherent 的投资展示了目标规模：科学家和设计师已经准备好了创新，但“没人为变化会如此之快做好准备”。
TSMC、GlobalFoundries 和 Tower 等传统厂商聚焦前道，制造硅光子晶圆；但 NVIDIA 采购的是最终芯片。激光器产能必须大幅扩张；设备厂商产出的设备不够；设备还需要改善吞吐量、效率和良率；衬底供应则“极其、极其紧缺”，原材料供应商只有少数几家。Van Hove 的判断是，行业必须把每一层的问题一路解决到最上游。
3. Themaa 正在搭建从实验室到晶圆厂的缺失桥梁
Van Hove 看到了大型晶圆厂负责前道、高校孵化公司停留在研究规模之间的缺口：市场上没有工业级参与者占据中间地带。Themaa 的答案是一家大型、专注光子业务的晶圆代工厂，目标是将数十年的欧洲研究产业化，而不是让商业化机会转移到其他地区。
团队的技术和运营背景本身就是这一判断的一部分。De Koninck 20年前进入硅光子领域，完成博士学位，先后在一家早期商业化初创公司和 imec 工作，5年前加入 NVIDIA，将其硅光子技术推进到可产品化的成熟度。长期从事投资、后来才转为经营者的 Van Hove 表示，他在3、4年前就看到了供应链缺口。
机会来自一次偶然事件：Themaa 在2023年至2024年规划期间，一家大型欧洲晶圆厂倒闭。收购这家工厂绕开了复杂基础设施审批、许可和建设通常需要的3至5年流程中的大部分环节，同时将公司置于距 imec 20公里、工业人才和光子学专业人才聚集的区域。
De Koninck 解释称，imec 30多年前从电子技术起步，后来成为硅光子的发源地之一，并发展为他所称的全球第一大先进电子技术研究中心。这个近在咫尺的研究引擎，支撑着 Themaa 从已被验证的科学成果迈向可重复的量产。
4. 客户提前拉动产能，Themaa 的路线图因此极其激进
Themaa 计划于2027年开始生产，并在2028年实现全面爬坡。Van Hove 表示，最大的挑战是客户施加的时间表；积极信号则是，客户已经提出超过公司“极其激进预测”的需求量。
项目执行仍取决于稀缺设备、性能更好的机器和衬底。缓冲来自分布在希腊、德国、西班牙、法国和瑞士的欧洲研究网络，各地分别处理不同的技术约束；Van Hove 认为，跨境协作如今比过去更具实质性。
招聘并没有成为预想中的障碍。imec 最资深的3名人士已加入 Themaa 董事会，公司还吸引了来自 NVIDIA、AWS、TSMC 和 Lam Research 的人才。Van Hove 希望在本土打造吸引人才的磁场，让欧洲专家留在本地建设，而不是出国或转行做银行家。
5. 欧洲有科学实力，Themaa 必须证明自己能以市场速度推进
美国头部潜在客户从4个方面质疑 Themaa：资金、人才、机器与衬底，以及欧洲能否满足其时间表。Van Hove 表示，Themaa 正有意回应欧洲“行动缓慢、受委员会牵制”的刻板印象，采用快速试错、果断决策、在没有“100%完美信息”时也采取行动的文化。
欧洲没有美国那样的数据中心网络，监管也更严格。Van Hove 认为，高能效基础设施提供了一个机会：欧洲可以建设自己的本土网络，其形态可能不同于美国的超大型云数据中心，而不是简单复制后者。
De Koninck 在 NVIDIA 学到的运营模式是：“组织架构只是一张纸”，任何能帮助项目的人都可以参与。离开他所称的“黄金宫殿”、登上“海盗船”，给了他一个机会：围绕 AI 的核心约束之一，打造世界级的欧洲制造能力。
Van Hove 更大的押注是，欧洲最终能够将自己的光子学领先地位商业化。他指出，根特大学声称在2000年至2015年全球4,000篇光子学同行评审白皮书中占据35%的份额，并以 ASML 证明欧洲过去确实打造过持久的工业冠军：“欧洲已经准备好迎来更多这样的故事。”
完整逐字稿
Yannick and Herwig, welcome to Sorcery, nStrike's annual gala. We're here in Monaco, and we're going to go deep into photonics and what you're building with Themaa. But before we get into that, we're going to go into a long conversation on photonics. Yannick, I'd love to start with your background, and then, Herwig, we'll get into yours. How did you get into photonics?
1. The Silicon Photonics Bet
I was one of these nerdy kids who locked himself up in his bedroom playing LEGO all day. So naturally, I went to engineering school. That was 20 years ago, and back then, silicon photonics was this niche, upcoming branch in engineering, with no prospect for any job at all. So I decided to jump on that ship.
I did my PhD in that field, and then, by that time, I had the opportunity to move to San Diego to join a startup company there that was trying to commercialize silicon photonics—one of the first companies to do that successfully. I moved back to imec, where I worked on laser integration for silicon photonics for a while, and then, 5 years ago, I joined NVIDIA, back when I still had to explain to people what NVIDIA actually was. We started from scratch and built the silicon photonics technology there into product-ready maturity. Now I'm starting at Themaa.
Herwig, how did you get involved?
I've been an investor most of my life and turned operator very late. I've been focused on deep tech for quite a while, so I got involved in the fascinating world of photonics almost a decade ago. I realized 3 or 4 years ago that the train was leaving the station, that there were some gaping holes in the supply chain, and that is what we're trying to solve with Themaa.
I realized that this visionary choice Yannick made 20 years ago to study the field of silicon photonics had put Belgium on the map and created a talent base that is broad and deep. I've spent my life selecting, honing, and nurturing talent, and so that's what we're going to hope to do with this.
2. AI Rebuilds Data Centers
Let's lay out the state of the market and how chips have gotten to this point, why copper is going out of date, and why there's a focus on photonics and lasers. Can you explain a little bit of that, and also why Europe?
I think the market is the part that really dragged me into it. We saw a much bigger opportunity than what most people in the field were seeing. That was because we were talking to the CTOs of data centers and realized what their need was and how photonics could help solve it.
We were listening very carefully when some of the gurus of the ecosystem, like Andy Bechtolsheim, were quoting numbers that far exceeded what even the most optimistic analysts were saying, and we felt that they were very well informed. That's what pushed us to do this.
To answer your question about where we are today, if you think about AI, you think about GPUs, right? You think about these compute engines that actually run the models. But the models have gotten so large that they no longer fit on a single GPU. So what we need to do is interconnect multiple GPUs to run these models.
Traditionally, that's been copper interconnects, but as these GPUs get faster, they need more data, and they need more data at a faster rate—
Mm.
—and copper is running out of steam. That's why we're transitioning to optical interconnects, which bring much higher data-transfer bandwidths. So, yeah, that's where we are today.
I recently interviewed Tony Kim of BlackRock in Paris at the RAISE AI Summit, and we were breaking down the entire industry. He says there's a $10 trillion rebuild happening, and one of the places most affected by—and most effective for—that spending is in these data centers, because the current infrastructure of a data center is not built for AI. It's completely inefficient. It takes a lot of water and a lot of energy, and the chips being used in those data centers are all outdated. They cannot handle the parameter-heavy models that are in the market.
I would love to get an understanding from one of you of the difference in architecture between those data centers and how they're transforming with this new build.
That's probably for you.
Yeah. If you compare a traditional hyperscale data center to a modern AI factory, one of the big differences is that there is way more traffic between different servers. If you think about your traditional Google search 5 years ago, basically, your traffic was routed to one computer that found the answer and sent that answer back.
Now, if you think about modern AI infrastructure, as I said before, the models no longer fit on a single GPU, so there's already GPU-to-GPU communication. But now, in the age of agentic AI, there's a model that's making a call to some tool, which is on a different server that's returning the results to the GPU.
There's a lot of traffic within the data center, which is very sensitive to latency—basically, the time between when all these different calculations can run. What that means in reality is that you need a much flatter network.
Hmm.
Instead of a really deep hierarchical tree of switches, you need a much flatter network. That means far more connections—direct connections between switches and GPUs. Basically, you need a much denser network than what we had in the traditional hyperscale.
If you're asking a CTO of a data center, “What keeps you awake at night?” the energy consumption is answer number 1, because it's existential to them. They don't know how to solve it. Their CEO is saying, “We need to grow, we need to grow,” and he's saying, “Okay, but where?” Nobody can deliver the grid.
Google negotiated for 3 years with Amsterdam and had to move its plans to Belgium because it simply can't get the grid. With the ambitious growth plans they have, that requires a much more fundamental solution that can scale to the same size as the problem. There are many data centers, and many more are being planned. Therefore, upgrading the grid or building new power plants cannot happen in the time required.
That's where every deep-tech paradigm shift needs a killer app. For photonics, more than 20 years in the making—or 30 years in the making on the science side—energy consumption by data centers and AI is that killer app.
Now we get to where we are today. That's great, but nobody was prepared for the speed of change. The speed of change happened in November '22. ChatGPT came out. It went viral like nothing we'd ever seen before. With that growth came this requirement, or this realization: “Okay, we need to switch to photonics.”
NVIDIA took the lead in that. They had been talking about photonics for a long time, but they said, “Okay, we need to make the switch happen.” In '24, they were even louder about it. In '25, they really pushed the market. In '26, they showed the market what they meant with their investments in Lumentum and Coherent. They've invested 4× the total market size in the first 6 months of this year.
Now everybody realizes, “Oh, the supply chain is not ready.” The innovation is ready. The photonic scientists have done a lot of work. The photonic designers have done a lot of work. But you need to work all the way back to the beginning, from the raw material through the tooling and the infrastructure required to build.
I think that's where the pain point sits today, and it's part of what we're trying to solve with Themaa.
How many photonic companies are out there? I interviewed the Lumentum CEO, Michael Hirson, and even just looking at that company, they are spilling over with demand. Their stock went up 1,000% in a year. It's insane. He had stepped in maybe a year and a half ago, so he really came in at an opportune time.
I'm curious: what does the state of the market look like for the companies that are existing, and how does it all work together?
3. The Photonics Supply Chain Gap
Again, what needs to be understood is the shift in the size of the market. Photonics for years was an academic endeavor, with some applications being prepared for in the supply chain at that scale. Now, all of a sudden, NVIDIA steps in and says, “No, no, guys, this is much, much bigger.”
Obviously, Lumentum and Coherent are the 2 leading companies in the West doing this. They're the only ones scaled at that size. If you look at the rest of the industry, the traditional players—the TSMCs, GlobalFoundries, and Tower Semiconductor of this world—have made a tremendous effort investing in the part that fits into their supply and production process, which is the front end: silicon photonics, or making the wafers.
But if you deliver a wafer to NVIDIA, that’s not what they ordered. They ordered a final chip, so there are a lot of other things required. Lasers, first and foremost, are required, which is what Lumentum and others are making. That’s going to have to scale by a lot, and that’s challenging because the throughput of the machines is not at all where it needs to be.
The efficiency and the yield are not what they’re supposed to be. The substrates—the raw material—are really, really cornered. The market is cornered: there are only a handful of suppliers. On top of all these problems, you now have the East-West divide, which is getting quite problematic.
In all of that, we think Europe is in a sweet spot. In Europe, you have industrial capacity that was available and industrial infrastructure that was available. That’s what we bought. You have photonic talent that has been trained for decades, and you have industrial talent that was available because all the big foundries closed. The combination of all of this, in addition to a desire for European sovereignty, puts us in an ideal spot to do this from Europe.
We have been building the concept of a pure-play photonic foundry at a much larger scale. We were looking around and seeing that all the big guys were doing the front end of the process, as I said, the silicon wafers. We were seeing a lot of innovation happening, but that was mostly at university spinout scale. There wasn’t really an industrial player in the middle.
As we were making plans in 2023 and 2024 to do this, I guess there was some serendipity involved in that a large-scale European foundry went out of business. We took the occasion and jumped on it. We bought it, and it delivered a very nice combination.
First, it was 20 kilometers away from imec, a world-leading research center with lots of talent. Second, there was lots of industrial talent, as I said, and infrastructure that was ready to go right at the time when everybody needed it. When you want to build a foundry of this size, you’re talking about a 3- to 5-year project. It’s between getting the permits and licenses and building everything. It’s not just clean rooms; the infrastructure is quite complex. Having all of it ready to go in 2024 was a serendipitous event, I’ll call it that.
That’s where we jumped on it, and then we started talking to both sides at the same time. When I say both sides, we started talking to the industrial talent and the photonic talent and said, “Look, Europe has successfully invested for 20 or 30 years in photonic innovation. This is the time for industrialization.” Yannick and others jumped on board instantly. So that’s where we are.
Yannick, from a technical standpoint, where are you at?
4. The Mission Behind The Move
I was in a great spot at NVIDIA, right? We were building the next generation of chips, and I wasn’t planning to go anywhere. But when I started to talk to Herwig and learned about the ambition of this enterprise and what he was planning to do, I realized that this was a once-in-a-lifetime opportunity for me to join a mission that’s going to solve some of the core problems that are slowing down probably one of the biggest, or the biggest, technological revolutions that I will witness in my lifetime. So I decided to leave the golden palace and jump onto the pirate ship—and set sail.
Selfish question: What is it like at NVIDIA? What have you learned there? What are the operating principles? What is it actually like?
It’s an incredible company. It feels like a startup. It’s very agile. The idea is to fail fast: we start building stuff, and if it doesn’t work, we move to something new. In my experience, it’s a very apolitical company. The org chart is just a piece of paper. If there’s a project and you can contribute, you contribute to that project. It doesn’t matter where you are in the org chart. It’s an incredible company.
And then, for people who don’t know what imec is, can you also explain imec and where it fits in the stack?
Sure.
And it’s not iMac, by the way.
No.
It’s not.
imec. Interuniversity Microelectronics Centre. That tells you something about how old this institution is. It started over 30 years ago and has been spearheading research, initially in electronics. For some time, it has also covered a much broader area; one of those areas is photonics. imec was one of the birthplaces of silicon photonics in the world and has grown to be the number-one research center for advanced electronics technologies in the world.
And where are you guys in terms of commercial progress to date? When are you going to start rolling out to your customers?
5. The European Photonics Buildout
Our biggest challenge is the timeline customers are putting in front of us. The good news is that the volumes they’re asking for even surpass our wildly ambitious projections, so that’s certainly very positive.
The constraints for us now are the same ones facing the whole market. Coherent, Lumentum, and others—everybody is facing the same constraints. The tool builders are not building anywhere near enough tools, and the tools themselves need to be further improved for throughput, efficiency, and yield. Even more problematic are the substrates, the raw material.
Here’s the good news: as I said, Europe has been investing successfully in innovation, and so some of these problems, even though they didn’t anticipate this shift, were being researched. Throughout Europe—Greece, Germany, Spain, France, and Switzerland, for sure—a pipeline has been built to tackle individual technical challenges around this.
To answer your question, our timeline is to start producing in 2027 and work toward full ramp-up by 2028. Our timelines are short, but Europe is wide awake and ready to go, and we feel it. That’s not just a slogan; we feel it every day. There’s a lot of collaboration going on, and we’re finally starting to really work across borders. I’m very optimistic about Europe playing a significant role, and Themaa playing a significant role, in this.
Has it been hard to hire?
That was the question I got most before I started, because I was all alone. Everybody said, “Where are you going to find the talent?” We are really enthusiastic about the people who are joining us. That’s true for the board, where three of the most senior people at imec have joined our board.
Obviously, we have people like Yannick from NVIDIA, but we also have people from AWS, TSMC, Lam Research—the top companies. There’s a lot of talent. Many of them are European, because what Yannick didn’t say is that he chose a niche field to study, but then you have a choice: you either go abroad to work, at least at the time, or you become a banker.
What we’re hoping to build with Themaa is a homegrown magnet for talent that becomes a self-sustaining ecosystem of innovation and industrial production. The answer to your question is no. We have been very fortunate with attracting talent.
So Europe’s been waiting for this?
Europe’s been waiting. Europe’s been waiting for this. If you read the European Council’s website on the European Chips Act II, they say that they really want to prioritize photonics. Within photonics, they really want to focus on pure-play industrialization. It’s almost like reading a definition of Themaa.
I think that has to do with European sovereignty, but it’s also because Europe realizes that it has what it takes not just to play, but actually to win in this game.
One of my favorite statistics comes from one of our board members, who was the visionary professor who founded the Photonics Department at the University of Ghent. Between 2000 and 2015, there were 4,000 peer-reviewed white papers on photonics globally. 35% of them came from the University of Ghent. I think, with the exception of Biscoff and chocolate, Belgium has no dominance like that in any other field.
6. Europe Reclaims The AI Opportunity
I'm curious to hear both of your answers to this question. I don't know if you travel to America much or what your exposure is like, but I'm very curious: what do you think the biggest misconceptions are about European AI?
Why don't you go first?
I have to think about that. What are the biggest misconceptions about European AI?
Maybe I'll flip it around. I can tell you what American tier-one potential customers asked us when they started speaking to us, which in and of itself was surprising. We were a startup with 2 people at that point in time, so they shouldn't have been talking to us, but they were.
What did they challenge us with? What did they ask us? They said, “One: can you find the funding? Two: can you find the talent? Three: can you find the machines and the substrates and everything else? And within all of that, how on earth are you going to respect the timelines we require from you?”
So the conception of Europe is that it's slow, that it's committee-driven decision-making.
I've heard about the summers here.
Yeah, exactly. So the conception of Europe is that it's slow, it's committee-driven decision-making. What we're definitely incorporating into Thema is, like Yannick said, this culture of fail fast, make decisions, drive forward, and don't expect 100% perfect information.
That may not be very European, but the world is globalized. We're awake, and we're going to take the best of American culture as well in the firm.
I was in California last week, and when I drove from San Francisco Airport to Santa Clara, along the highway, there were all these billboards on AI stuff. So you can feel there's incredible excitement about AI.
I don't think many people who live there like those billboards.
That's another thing, but you can feel the excitement, right?
Yeah.
I think we have a very similar excitement here in Europe. We may not be as vocal about it, but we're going to turn that excitement into a reality.
Is there as much hate for data centers here as there is in America? I mean, it's a huge media propaganda thing—I’ll be clear about that—in America, but I'm really curious how it's playing out here.
Europe doesn't have the same network of data centers as the US has. The regulations here are a lot stricter. So I would almost flip it around: that's the opportunity for us, right?
If we can help build the infrastructure that makes this whole business a lot more energy-efficient, then hopefully we will also start building our own homegrown data center network in Europe, which may look very different from the mega-hyperscalers that the US has built. But that's part of the problem we're trying to solve.
Europe doesn't have the same aversion to it because it simply doesn't have the problem; it was regulated before the problem arose.
7. Photonics Powers AI Infrastructure
What timelines are you thinking about? I ask this to people all the time, and I talk to aerospace companies. They're like, “We're in dog years. A month is like 10 years for us.” But then you have other companies where it's a year. Anyway, what timeline are you thinking about, and what are you most looking forward to within that timeline?
We're getting ready for production in 2027. What I'm most looking forward to is proving the incredible gaps that we're going to cover from lab to fab, and we have the right team on board to do that.
There is still a lot of ground to cover. I wouldn't call it low-hanging fruit, but the experience of the team combined with the photonic talent is going to allow us to beat expectations on that. That's what I'm looking forward to. The bar is high, but I think we can beat expectations.
Yannick, what are you most looking forward to?
Just getting started and seeing how the community reacts, then adjusting to that and building some world-class manufacturing capability here in Europe.
I mean, you're coming from a very competitive company, so do you bring that competition with you? Is that why this is so enticing?
I think so, yeah. I like a challenge. This is obviously a very competitive field right now. Everybody sees, or is starting to see, this opportunity. So we'll have to be fast, we'll have to be precise, but I'm very much looking forward to that.
Are there a lot of companies spinning out of NVIDIA in Europe right now?
Not that I'm aware of.
No?
No.
Okay. Unicorn. Awesome. As we close out, I have to ask you one last question. It's the most difficult question. You can take time to think about it if you need to, but what is your hottest take right now?
Hottest take.
It could be that Europe is underrated, it could be that photonics are underrated, it could be—
I think both of these are true, actually. People are waking up to the potential of photonics, but it's just the start of a much larger manufacturing ecosystem that's going to exist. That's one.
And yes, Europe is underrated. The innovation that has occurred here has not been commercialized in Europe. Whatever was good enough was either taken away to Taiwan or to the US. So to build it here, Europe has done this once. It's a long time ago, but it's been a 40-year success story, and it's called ASML.
That's been incredible, and I think Europe is ready for more stories like that. It's underrated how ready Europe is for that.
And I think, for me, the hot take is that if you look at an AI factory today, half of that is GPUs, but the other half is the technology to interconnect these GPUs. There's a huge opportunity for optics, for photonics there, and we have to tap into that.
Amazing. Great place to end it. Well, Yannick, Herwig, thank you so much for a nice little deep dive on photonics and Veoma.