### AI 칩 개발: 시장 차별화를 위한 전략

### 경제적인 AI 솔루션 구축: 미래를 위한 비전
* 💡 이 회사는 강력한 팀을 구성하고 제품을 출시하는 동시에 개발자와 초기 고객을 참여시키는 데 집중하면서 **가속화된 칩을 위한 저비용 운영**을 목표로 하고 있습니다.
* 🛠 소프트웨어 엔지니어를 끌어들이는 오픈 소스 AI 컴파일러 스택을 통해 투명성을 강조하는 한편, **IP 라이선스를 제공**하고 **합리적인 가격의 개발 시스템**을 제공하는 등 독특한 판매 포인트를 가지고 있습니다.
* 💰 특히 비용이 많이 들고 제한이 있는 **고대역폭 메모리 칩**을 사용하지 않음으로써 비용을 절감할 계획입니다.
* 📊 내년 매출 **2억 달러를 목표**로 이미 **1억 5천만 달러를 예약**한 이 회사는 차별화를 추구하는 소규모 업체를 타깃으로 삼고 있는 등 시장 잠재력은 방대합니다.
* 🌐 Jim은 향후 **10년 안에 AI가 컴퓨팅을 지배할 것**이라고 믿으며, 하드웨어와 소프트웨어 모두에서 상당한 변화를 앞두고 있는 이 산업은 이제 막 시작 단계에 있다고 강조합니다.
* 🚀 그는 성공적인 기업은 최고의 기술을 만드는 데 집중한다며 **실행보다 훌륭한 제품**을 만드는 것이 중요하다고 강조합니다.
* ⚖ 이 회사는 규정 준수에 전념하면서 **미국과 중국의 무역 제한을 신중하게 탐색**하는 한편, 변화하는 환경에서 가능한 방대한 비즈니스 기회를 모색하고 있습니다.

Jim은 소규모 업체를 위한 합리적인 가격의 오픈소스 솔루션으로 AI 칩 개발에 혁명을 일으키고자 합니다.

The money that you're raising, what is
it that you're hoping to do? You're wanting to make a small, cheaper
operation, an option out there for accelerated chips?
Yeah, that's right. We're building out the team.
We're building out our products.

We're starting to ship our products.
And our real mission right now is to bring on developers and early customers.
Let's talk about the customers and what you offer that isn't in the market at
the moment. Jim, what's your selling point?
Well, we have a couple of things.

We licensed our IP and we're building a
business on that for both risk Fi's CPU and I, we're selling low cost
development systems. Like, I'm really interested in hitting a
much lower cost point. Mm hmm.
And then we open sourced our whole AI

compiler stack.
And people like the transparency. They can see exactly what the technology
is doing. They actually it's turned into a great
recruiting vehicle for software engineers.
Interesting. So let's just go first to the price
point.

Just a cheaper offering.
A lot of that is done by having less pricey components.
For example, Nvidia's chips are incredibly expensive.
We know people pay it because of what they offer the full stack in many ways,
but they struggle from a choke

perspective on high bandwidth memory
chips in particular. You don't need them.
Why do you need them? Why do you want them?
Well, it's partly about choices. So there's two choices.
One is if you pick a really expensive

memory technology, you'll never get the
price down, right? And so you need to then make technology
decisions to lower your costs. So we have a in a tensor processor
architecture. We keep the data on chip more.
We have a software stack that's very

good at partitioning A.I.
workloads across multiple chips, and then they can send information and
compute packets to each other without having to go through memory.
And that lowers the memory bandwidth required and also lets it scale more
naturally across many chips.

So I'm a new investor.
You brought me on board. I'm interested, Basil, Samsung.
But you're saying, look, I'm going to do this with more open architecture, with
cheaper offerings. How much of the market should you have
to get?

What sort of price point are we looking
at for your chip offering? Yeah.
So this is really wild. The market is so big.
Like the percent we have to get is actually pretty small to be a real value
proposition. You know, next year we'll die and go to
heaven if we make $200 million.

We have already booked 150 million in
revenue. I'm really interested in finding small
players, early players. People want to own their technology.
People are excited about our software and hardware architecture.
And you know, I want to grow organically

from there.
I think some people will focus on the big guys first.
That's not how new technology introductions go.
You focus on smaller players, people who need a differentiation, and that's what
we're really that's our mission.

So what therefore is the mission
building in in terms of long term value of generative AI, of AI applications of
do you think that we are in some sort of hype cycle?
Do you think that actually this market is as big as many are factoring in?
Well, first of all, we are definitely in

the hype cycle.
There is going to be multiple. And people say, well, Jim, you saying
it's going to go down next year and that's going to go up and down, low pass
filter, all that stuff. Right.
What's really going to happen is AI

is going to dominate computing over the
next ten years and we're just starting and the hardware is going to change a
lot. The software is going to change a lot.
Like one day you read in the papers, alums can do anything, and the next day
you read they've hit a limit. 

Ignore all that stuff.
We're just starting this. There's going to be massive
transformations. So what I really want to start to have
is great computer designers, general purpose CPUs, AI
processors, chip design systems, design, 

software design, and then the open
source stuff is really interesting because that engages a community.
People are really passionate about that to contribute back.
This is why you're thinking about your own talent team.
You've managed talent, as we said, sort

of across the board.
You began like you were integral to the chip design process in the US.
When you think back to DC and then the work that you did at AMD.
Moving to Apple, you were Intel and I just have to go there on this day that
Pat Gelsinger hands in his room, hands

his hat on off, he's going to be
resigning. Many would say it's because they didn't
get into AI quickly enough. That was prior to him.
What do you make of the transition? Well, it's hard to say.
So I like Pat personally.

I've met him a number of times.
He's a smart guy. He really cares.
I think the market focuses too much on stuff like execution and transition.
The way you make money, get customers is build great products.
Like Steve Jobs was the best at this.

Focus on the product, make the best
product you possibly can. And I think they need somebody way more
hands-on, really focus on product and they need to stop talking about
execution. All the nonsense, right?
When they make the best product, the

best fab, the best chip, the best CPU
accelerator, they'll do fine. Jim.
When you're thinking about all of the areas you want to own, how much does
U.S. and China factor into your growth story?
So, first of all.

We are building open source software
RISC-V based technology. We think that's an open platform.
There's a lot of people interested in that.
At the same time, you know, there are, you know, restrictions by the US
government and trade rules.

We have really good lawyers and trade
compliance people. We're very closely with those guys.
We want to do the right thing both for the business and for, you know, being
compliant to what's going on. And it's a changing world and we're very
involved in that short run.

Our business possibilities are really big. And, you know, if China plays in that, that'll be interesting. If it doesn't for various reasons, that's fine.

Tenstorrent's Keller: We're in an AI Hype Cycle

Welcome to the Bloomberg Technology YouTube channel. Subscribe for original series and mini-docs on game-changing technologies, profiles on innovators, live conversations with experts, and breaking technology news & analysis from Bloomberg TV.

Bloomberg Technology

Focus on building a team, products, and shipping to early customers and developers.

Selling point: low-cost development systems and open-sourced AI compiler stack.

Licensed IP used for both CPU and AI, aiming for a much lower cost point.

Open-sourced AI compiler stack improves transparency and helps recruit engineers.

Cost Reduction Strategies and Technological Choices

Lower costs achieved through less expensive components, avoiding high bandwidth memory chips.

Choosing less expensive memory technology to lower overall costs.

Utilizing a tensor processor architecture to keep data on-chip and optimize software for workload partitioning across multiple chips.

This reduces memory bandwidth requirements and enables better scalability across multiple chips.

Targeting a small market share to achieve a significant value proposition.

Projecting $200 million in revenue next year, having already booked $150 million.

Focusing on small, early players who want to own their technology and are excited about the software and hardware architecture.

Organic growth strategy targeting smaller players needing differentiation.

Acknowledging the current hype cycle but emphasizing AI's long-term dominance in computing.

Expecting significant transformations in hardware and software over the next ten years.

Ignoring short-term fluctuations and focusing on long-term growth and innovation.

Emphasis on building a strong team of computer designers and leveraging open-source contributions.

Discussion of the speaker's experience in chip design and leadership roles at various companies.

Commentary on Pat Gelsinger's resignation from Intel, suggesting a need for stronger product focus.

Emphasis on the importance of creating great products over focusing on execution and transitions.

Advocating for a hands-on approach to product development and minimizing distractions.

Belief that creating superior products will lead to success.

Geopolitical Considerations and Compliance

Building open-source, RISC-V based technology for a global market.

Acknowledging US government restrictions and trade rules.

Maintaining strict compliance with trade regulations and working closely with legal experts.

Adapting to a changing global landscape and remaining open to opportunities in various markets, including China.