<div id="readability-page-1" class="page"><div>//php echo do_shortcode('[responsivevoice_button voice=&#34;US English Male&#34; buttontext=&#34;Listen to Post&#34;]') ?&gt; AI&amp; (“AI And”), a new vertically-integrated Japanese AI company from former Tenstorrent and Lenovo executives David Bennett and Shimpei Hara, has raised $50 million in seed funding and $2 billion in infrastructure capital.AI&amp; wants to tackle every part of the AI stack from the data center upwards, including building out Japanese data center infrastructure as well as developing orchestration/cluster management software for heterogeneous data centers, plus AI models, agents, and applications. At GTC 2026 last week, Nvidia CEO Jensen Huang postulated that all employees of all companies will be running 100 AI agents each in the near future. If scaled to Japan’s population, with each agent using one million tokens per day, total power demand would reach a terawatt, Shimpei Hara, AI&amp; president and cofounder, told EE Times. “If this is where we’re heading, who is going to be the server of AI in Japan?” he said. “And how can you serve Japan without breaking the bank and without needing a terawatt of power, because that’s definitely not the direction our government is heading towards.” AI adoption at such scale will require a vertically-integrated approach to be economical and efficient, Hara said. This will mean AI&amp; can optimize at every layer of the stack to improve the speed and quality of tokens while enabling some flexibility in pricing. “Others are trying to carve out a technical specialty within their specific domain and make a business out of it, so there’s margin stacking, to some extent…which results in a highly inflated end token cost and AI deployment cost,” he said. “We believe that having ownership across the stack and optimizing through that is actually going to help us reduce the cost of serving overall, specific to the Japan market.”<h3>Japanese market</h3>The company will focus on Japan, with a view of a global rollout further down the line, AI&amp; CEO and cofounder David Bennett told EE Times.“[The Japanese market] is driven by AWS and the big hyperscalers,” Bennett said. “We see a real opportunity because there are complexity and price issues dealing with the big guys. Whether there are security or privacy reasons, or other workload-specific things, there’s certainly a need for players like us.”The real problem is how to get enterprises up and running with AI, he said, since they need to engage higher up the stack than the hardware level. In this case, great technology isn’t enough—it has to be easy to use.There’s a big demand for data sovereignty, Bennett added.“People want to own their data, they want open models, they also want to know that if they’re putting their data into [online AI chatbots] where that data is going,” he said. “We’ll be able to say it’s staying in our data centers in Japan.”AI&amp; plans to prove its concept in Japan before looking at Southeast Asia, Europe, and beyond.<h3>Heterogeneous hardware</h3>The infrastructure and staff of Japanese AI cloud provider Unsung Fields have been folded into AI&amp; (Unsung Fields has ceased its operations). This means AI&amp; already has two Japanese data centers up and running, offering more than 1,000 GPUs plus a cluster of Tenstorrent hardware.The heterogeneity of its data centers is a key part of the company’s strategy, Bennett said.“Being able to send the right workload to the right hardware, and frankly, the right kind of token to the right user—we really believe in the heterogeneous aspect,” he said. “Whether it’s disaggregation strategies or not having to be limited in your supply chain to a single vendor, [heterogeneity] is obviously important.”The company already has a significant amount of Nvidia GPUs and is planning to work with AMD hardware, too, Hara said. The company expects to open another data center in Japan within a month to help serve its 80 existing customers.“Our first heterogeneous experiments are around how to get efficient token throughput gains by disaggregating across AMD and Nvidia,” he said. “We think that’s a really interesting thing to continue to push, [because] even a 1.5 to 2× in token throughput is [worth it].”As well as its GPU clusters, AI&amp; is already generating revenues from Unsung Fields’ Tenstorrent hardware.“AI&amp; has probably the largest installation of Tenstorrent hardware anywhere, and we’re really looking forward to what they’re doing with Blackhole,” Bennett said.Acknowledging that heterogeneity can add complexity to the stack, the company will start simple, Bennett said. “We’re not creating complexity in our tech stack for the fun of it,” he said. “It’s quite simple today, we can route the right model to the right place, that’s super easy. We’re going to be very selective in how we use our heterogeneity and how complex we get. If there’s really no value, we’re not going to do it.”<h3>Japanese AI lab</h3>AI&amp; also plans to become Japan’s premier AI lab.The big AI labs will continue to build foundation models, but there’s also space for development of more application-specific models, he said.“If you can put the right amount of effort into it, the right amount of data in pre-training, post-training and RL [reinforcement learning], and then you build the right harness, you can provide some workloads, experiences and use cases that outperform the foundation models on very specific tasks,” he said. “In markets that are under-served because they’re not English-based, or they have different needs, they can be better served when you have a team that’s designing and building models, including foundation level models that cater to that market.”AI&amp; also plans to become an incubator for Japanese AI startups, enabled by being vertically integrated, Bennett said.“There’s a lot of AI startups in Japan that have great ideas and need a ton of compute,” he said. “I think it’s a perfect match for us to incubate these startups, and we can do some really interesting things around…business models where we can both benefit.”Ultimately, there are a lot of focused AI researchers and developers in Japan who want to do something in Japan, for Japan, and they need somewhere to do that, Bennett said.“Japan has a lot of talent and they need a place to go,” Bennett said. “We want to be that place to go.”<hr/><h5>See also:</h5>Layoffs At Tenstorrent As Startup Pivots Towards Developer SalesCost-Efficient AI at Scale is a Software ProblemJim Keller: ‘Whatever Nvidia Does, We’ll Do The Opposite’ </div></div>

LNVGY

4488

4388

DTCR

SRVR

IDGT

CLOU

AI& is a new AI company in Japan founded by former Tenstorrent and Lenovo executives, raising $50 million in seed funding and $2 billion in infrastructure capital. The company aims to build a vertically-integrated AI stack, focusing on Japanese data center infrastructure and AI applications. AI& plans to address the complexities and costs associated with existing providers like AWS, emphasizing data sovereignty and ease of use for enterprises. With two data centers operational and plans for expansion, AI& is set to optimize AI deployment in Japan before considering global markets.

Ex-Tenstorrent Execs Start Cloud Provider, AI Lab in Japan