"SoftBank acquires UK AI chipmaker Graphcore"

 From TechCrunch, July 11:

Terms of the deal were not disclosed, but Graphcore CEO Nigel Toon says it’s a positive outcome (for most). Full regulatory approval has been granted already, meaning this is final. 

U.K. chip company Graphcore has been formally acquired by Japan’s SoftBank.

Rumors of the deal have abounded for some time, but protracted negotiations and regulatory approvals have meant neither company has confirmed anything until now. Even today, the company wouldn’t confirm the one thing most people will be wondering: How much does Japanese multinational SoftBank value a startup touted as a potential rival to the mighty Nvidia in the AI chip space?

While the figure of $500 million has been bandied around in various reports for months, in a press briefing early Thursday morning, Graphcore co-founder and CEO Nigel Toon remained coy on the details. “We have agreed with SoftBank that we’re not going into the details of the deal; whether anything comes out in the future, we’ll see,” Toon said.

Toon did say, however, that the $500 million figure was inaccurate. Make of that what you will.

When the chips are down

Founded out of Bristol in 2016, Graphcore has devised a new kind of processor dubbed an “intelligence processing unit” (IPU), distinct to the kinds of graphics processing units (GPU) developed by the likes of Nvidia. While both accelerate computation, IPUs have a different architecture designed from the ground-up for AI workloads. Graphcore pitches its chips as a more efficient alternative to GPUs, with a focus on supporting large-scale parallel processing and executing complex machine learning models, where the model and data are tightly coupled.

Graphcore had raised around $700 million since its inception, reaching a valuation just shy of $3 billion in late 2020. With big-name corporate and institutional investors like Microsoft and Sequoia, and angels such as DeepMind’s Demis Hassabis and OpenAI co-founder Greg Brockman, hopes were high that Graphcore could be an AI beacon in the U.K. or even all of Europe. But AI hardware is a resource-intensive business, and Graphcore ultimately couldn’t hit the giddy heights many had hoped it could reach. It lost out on potential lucrative cloud deals with Microsoft, while the U.K.’s own government ignored Graphcore (despite a public plea from Toon himself) for its new “exascale” computer plans last year.

Graphcore has not had the best of times of late, compounded last year by its forced China exit due to U.S. export rules.

With losses widening and Graphcore approaching four years since its last capital injection, it was increasingly clear that something had to happen somewhere — and an acquisition always seemed the most likely outcome, particularly at a time when demand for AI hardware is at fever pitch.

SoftBank, for its part, is no stranger to U.K. semiconductor companies, having previously acquired Arm for £24 billion ($31 billion) and then retained a stake as it spun Arm out as a $55 billion publicly traded company last year. Arm is now worth close to $200 billion — a sign, perhaps, that SoftBank might not be the worst bedfellow for Graphcore, as the well-financed Japanese powerhouse seeks to bolster its AI aspirations with everything from data centers and robotics, to the semiconductors needed to power the AI revolution....

....MUCH MORE

Most recently (May 8):
Chips: "SoftBank Is Said in Talks to Buy Troubled AI Chip Firm Graphcore"
In 2017 it was looking so promising: "Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC).

"Inside the UK unicorn that's about to become the Intel of AI"

In November 2016 we headlined a post Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA).
Bristol?...

Graphcore was the "Office in Bristol".
A year later: "Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC):
Huh.
 
From Wired, August 27:

Bristol-based Graphcore's "intelligence processing unit" aims to do for AI what the graphics processing unit did for computing

In September 2015, hardware veterans Nigel Toon and Simon Knowles were doing the rounds of venture capital offices in Silicon Valley and London, touting their latest startup. The pair had a dazzling track record – among other achievements they’d sold their previous semiconductor company Icera to NVIDIA for $435 million (£346 million) four years earlier. And their vision for Graphcore – a new Bristol-based venture – was bold: they were building a new generation of microchips known as intelligence processing units (IPUs), designed for the rapidly approaching artificial intelligence age.

Yet early reactions to their pitch for series A financing were distinctly muted. “In many cases we were laughed out of court,” recalls Toon, Graphcore’s CEO. 

Typically, Toon says, they’d find a partner in a VC firm who was excited by what they were doing. “But then they’d go to their partner meeting, where the first question would be: ‘What’s AI?’ It’s stunning to think that was a conversation that was happening [as recently as] 2015.” From there, it was an uphill struggle. “Even if they got the fact that AI might be interesting, they’d then say: ‘Your business model is to build a chip for this AI thing? Well, nobody’s made money from chip investments in the last 10 years.’”

Toon, who is 55 and has the mellifluous voice of an old-school BBC continuity announcer, says that chip development, in the eyes of most investors at the time, was considered highly capital intensive, with returns failing to justify the upfront financing required. “It’s not more capital intensive than software,” says Knowles, Graphcore’s co-founder and CTO. “But software has this joyful property that you can try it out in small scale first, whereas with a chip you’re all in. If it doesn’t work, you’ve spent all your money.” 

That was 2015. Fast forward to today and, of course, AI hardware is a white-hot category for investors, with VC funding for US AI companies jumping by 72 per cent in 2018 to a record $9.3 billion (£7.4 billion), a fifth straight year of growth, according to a report by CB Insights and PwC.

What changed over those three years? Toon points to two things. First, in 2016 traditional chip giant Intel acquired an AI software and hardware startup called Nervana for $350 million (£280 million), raising eyebrows all over the Valley. Second, Google announced it was going to build its own chips – evidence, Toon says, that existing chips weren’t up to the task. 

Knowles describes the impact of Google’s decision as “seismic”. The fact that Google thought AI was going to be a sufficiently big deal to justify the pain and expense of building its own chip team helped make the Graphcore founders’ case for them. He and Toon had been arguing that it was worth digging deep financially to develop new processor hardware because existing graphics processing units (GPUs) – used, for example, in mobile phones, games consoles and personal computers – weren’t designed for AI workloads such as machine learning and deep learning. 

By then, their startup was already ahead of the pack in developing a new processor architecture. Soon top-tier investors – including Atomico, one of Europe’s best-known VCs – were beating a path to their door. Atomico, which went on to lead Graphcore’s $30 million (£24 million) Series B round in July 2017, was followed six months later by one of the Valley’s biggest guns, Sequoia Capital. At the time, Graphcore, having recently closed its Series B, didn’t need investment – but the west coast investor wasn't taking "No thanks" for an answer. “They came to see us here in Bristol and said, ‘No, you don’t understand, we want to invest in your business,’” laughs Toon. “So we work out terms and they invest $50m into the company. And that’s one of the very few investments they’ve made in the UK, because they’ve got so much opportunity on their doorstep.” 

Sequoia partner Matt Miller, who now sits on Graphcore’s board, admits he was somewhat bemused to find himself chasing down a company based in Bristol. “We knew there was an opportunity for a new architecture that would be designed from the ground up that could massively accelerate our entry into this AI age, and we were trying to landscape all of these companies in China, the US and Europe,” he says. “But our references were all pointing to this one company in Bristol, whom we hadn’t met yet.” 

A roar of laughter distorts the line from the Valley. “Lemme tell you, if you’d asked me a month prior if I’d ever [sit on] a board in Bristol I’d have said ‘No way!’ It’s not your typical destination on your tour of Europe. But to be honest, it’s been surprising for us in the Bay Area because the quality of talent in the UK, and particularly in Bristol in the semiconductor space, is very strong. The team they’ve been able to build there is on a par with the best in the world.”

Following a $200 million (£160 million) Series D round in December 2018, Graphcore was most recently valued at $1.7 billion (£1.36 billion), with investors, innovators and large corporates now seemingly convinced it will be the company to power the AI era in much the same way as Cambridge-born chip giant ARM dominated mobile devices, shipping over 130 billion chips and reaching 70 per cent of the global population. The opportunity at stake is nothing less than the future of AI, with applications ranging from medical advances to autonomous vehicles, space exploration and just about everything in between....MUCH MORE 

If interested, see also:

Aug 26 

"A look at Europe's 5 most valuable AI startups"

June 13 

"New Artificial Intelligence Chips Lean Toward the Edge" (NVDA)

April 22

Graphcore CEO Takes A Shot at NVIDIA, Talks His Book, Makes a Good Point (NVDA)

Dec. 26, 2018

"7 Predictions for AI in 2019"

Dec. 19 

British AI Startup Graphcore Raises $200 Million From BMW, Microsoft

May 2018 

Top 10 British Artificial Intelligence Startups

January 2018 
"Can Chinese AI Chip Makers Compete with Nvidia?" (NVDA)
 

And many more, use the 'Search blog box if interested

British AI Startup Graphcore Raises $200 Million From BMW, Microsoft

Way back in 2016 we posted "Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA)":

Bristol?
From HPCwire:

November 1, 2016

Graphcore emerged from stealth mode today with news of a $30 million Series A round to help finance ongoing development of its machine learning (ML) and deep learning acceleration solutions, including a PCIe card that plugs directly into a server’s bus....

And here we are. From Bloomberg via Yahoo Finance:

A U.K. startup that designs semiconductors used for artificial intelligence applications has raised $200 million from investors including BMW AG and Microsoft Corp.

Graphcore Ltd. is one of a number of companies trying to design a new class of chips that will be better at crunching the vast amount of data needed to make computers smarter. They argue that the processors that have defined the PC age -- made by Intel Corp. and Nvidia Corp. -- are not suited for this task, and more tailored solutions are needed to achieve the kind of speed up required.

The funding round was led by U.K. venture capital firm Atomico, and valued Graphcore at $1.7 billion, the company said Tuesday. Existing investors Dell Technologies Inc. and Robert Bosch Venture Capital also participated in the round.

The billion-dollar-plus valuation for the chip company comes at a time when the semiconductor industry is suffering from investor skepticism, but AI is getting increased attention. The U.K. company’s chips are specifically designed to accelerate machine learning -- the process by which predictive computer algorithms improve from digesting large amounts of data. Graphcore designs chips for power-intensive tasks such AI in cloud computing, enterprise services and automotive systems.

Graphcore, headquartered in Bristol, England, will also be opening up new offices in Beijing and in Hsinchu, Taiwan. With plans to eventually IPO, it is currently rolling out its first run of chips and is working on its next product. Initial clients include Dell and Samsung Electronics Co., said Graphcore co-founder Nigel Toon. The company is targeting $50 million in revenue in 2019.

Graphcore’s technology "is well-suited for a wide variety of applications from intelligent voice assistants to self-driving vehicles,” Tobias Jahn, principal at BMW i Ventures, said in a statement.

Last year the company raised $50 million from investors including U.S. venture capital firm Sequoia Capital. Previous investors include AI luminaries Hermann Hauser, co-founder of Arm Holdings Plc, and Demis Hassabis, co-founder of Google’s DeepMind....

If interested see also: 

Nov. 2017

"Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC)

May 2018
Top 10 British Artificial Intelligence Startups

The GOOG was the "lab in Shenzhen":  AI: "Google moves into Shenzhen in latest China expansion" (GOOG; NVDA)

I am not at liberty to discuss the basement in Moscow.

Graphcore CEO Takes A Shot at NVIDIA, Talks His Book, Makes a Good Point (NVDA)

In November 2016 we headlined a post Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA).
Bristol?...

Graphcore was the "Office in Bristol".
A year later: "Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC):
Huh.
Sometimes you get lucky...

From EE Times:

GPUs Holding Back AI Innovation
GPUs are widely used to accelerate AI computing, but are the limitations of GPU technology slowing down innovation in the development of neural networks?

In a recent interview with EETimes (Graphcore CEO Touts 'Most Complex Processor' Ever), Nigel Toon, CEO of Graphcore, explained that while GPUs are good at running convolutional neural networks (CNNs), they are not suitable for running the more complex types of neural network needed for reinforcement learning and other futuristic techniques.

“A GPU is a pretty good solution — if all you’re doing is basic, feed-forward CNNs. The problem comes when you start to have more complex neural networks. Rather than just doing it a layer at a time, I want to be able to go through some layers then feed back, and I want to be able to store information on the side which I can use as context information as I look at the next data. And if my data is changing — so, rather than it being millions of static images that I can feed in in parallel — if it's video and I'm interested in sequential frames, it's much harder to feed that in in parallel and take advantage of the wide SIMD paths in a GPU,” he said.

Visionary Approval
As part of our longer conversation, Toon noted that Graphcore has captured the interest of such AI visionaries as Demis Hassabis, a founder of DeepMind, and the founders of OpenAI, including Ilya Sutskever, along with many other leading researchers in machine learning. Graphcore worked with these researchers to design the IPU architecture based on the kinds of problems that they want to solve.

“All the innovators we spoke to said [using GPUs] is holding them back from new innovations,” he said. “If you look at the types of models that people are working on, they are primarily working on forms of convolutional neural networks because recurrent neural networks and other kinds of structures, [such as] reinforcement learning, don't map well to GPUs. Areas of research are being held back because there isn't a good enough hardware platform, and that's why we're trying to bring [IPUs] to market.”

Processor Development
Toon points to the development of ASIC-type accelerators that are built to accelerate specific neural networks, as well as increased interest in FPGA solutions for AI, as proof that GPUs can’t do the job well enough. There is a need, he says, for an easy-to-use processor that is designed from the ground up, specifically for machine intelligence.

“What you need to do is to extract parallelism in many different dimensions. So, rather than having an SIMD processor, what we need is a multiple-instruction, multiple-data machine,” he said. “We need to solve the problems of: ‘How can we access the memory in real time, during the compute?’ ‘How can I take pieces of data from here and there, gather that together, do the compute, and then scatter the answer back somewhere else?’ These are all the things that we have been solving with the IPU processor....MORE

If you follow the link there are three references to the entire interview. If you don't follow the link here's "Graphcore CEO Touts 'Most Complex Processor' Ever".

If interested see also:
Top 10 British Artificial Intelligence Startups

British AI Startup Graphcore Raises $200 Million From BMW, Microsoft

And related (and the reason the two types of chips are bolded above):
Chips: The Accelerator Wall—A New Problem for a Post-Moore’s Law World (GPU; ASIC; FPGA)
Watch Out Nvidia, Xilinx Is Performing (reports, beats, pops) XLNX; NVDA

Xilinx with their field-programmable gate array approach versus Nvidia's more generalist chips is an example of the type of competition experts were predicting NVDA would be facing, some over two years ago, links after the jump...

A Dip Into Chips: "AI Chip Architectures Race To The Edge"
"Why Alibaba is betting big on AI chips and quantum computing"
Hot Chips 2018 Symposium on High Performance Chips
Chips: "A Rogues Gallery of Post-Moore’s Law Options"
"Top-Rated Chipmaker Xilinx Gets Big Price-Target Hike On 5G Prospects" (XLNX; NVDA)

"Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC)

Huh.
Sometimes you get lucky.

November 21, 2016
Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA)

From the high performance computing geeks at HPC Wire:
November 13, 2017

BRISTOL, England, Nov. 13, 2017 — Graphcore has today announced a $50 million Series C funding round by Sequoia Capital as the machine intelligence company prepares to ship its first Intelligence Processing Unit (IPU) products to early access customers at the start of 2018.

The Series C round enables Graphcore to significantly accelerate growth to meet the expected global demand for its machine intelligence processor. The funding will be dedicated to scaling up production, building a community of developers around the Poplar software platform, driving Graphcore’s extended product roadmap, and investing in its Palo Alto-based US team to help support customers.

Nigel Toon, CEO at Graphcore said: “Efficient AI processing power is rapidly becoming the most sought-after resource in the technological world. We believe our IPU technology will become the worldwide standard for machine intelligence compute. The performance of Graphcore’s processor, compared to other accelerators, is going to be transformative, whether you are a medical researcher, roboticist, online marketplace, social network or building autonomous vehicles....MORE

...Last month, Graphcore shared preliminary benchmarks demonstrating that its IPU can improve performance of machine intelligence training and inference workloads by 10x to 100x compared with current hardware. One test showed that a developer could use eight IPU PCIe cards to run a training model in the same amount of time as 128 GPU cards....

Chips: " Graphcore Uses TSMC 3D Chip Tech to Speed AI by 40% Unveils plan for $120-million “brain-scale” supercomputers in 2024"

In November 2016 we headlined a post Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA).
Bristol?...

Graphcore was the "Office in Bristol".  

From IEEE Spectrum:

U.K.-based AI computer company Graphcore made a significant boost to its computers’ performance without changing much of anything about its specialized AI processor cores. The secret was to use TSMC’s wafer-on-wafer 3D integration technology during manufacture to attach a power-delivery chip to Graphcore’s AI processor.

The new combined chip, called Bow, for a district in London, is the first on the market to use wafer-on-wafer bonding, say Graphcore executives. The addition of the power-delivery silicon means Bow can run faster—1.85 gigahertz versus 1.35 GHz—and at lower voltage than its predecessor. That translates to computers that train neural nets up to 40 percent faster with as much as 16 percent less energy compared to its previous generation. Importantly, users get this improvement with no change to their software at all.

“We are entering an era of advanced packaging in which multiple silicon die are going to be assembled together to supplement the performance advantages we can get from increasing progress along an ever-slowing Moore’s Law path,” says Simon Knowles, Graphcore chief technical officer and cofounder. Both Bow and its predecessor the Colossus MK2 were made using the same manufacturing technology, TSMC’s N7....

....MUCH MORE

Previously:
August 2021
"NVIDIA and the battle for the future of AI chips"
June 2020
"AI chips in 2020: Nvidia and the challengers"
August 2019
"Inside the UK unicorn that's about to become the Intel of AI"
April 2019
Graphcore CEO Takes A Shot at NVIDIA, Talks His Book, Makes a Good Point (NVDA)

And many more. Use the 'search blog' box upper left, if interested.

Chips: "SoftBank Is Said in Talks to Buy Troubled AI Chip Firm Graphcore"

In 2017 it was looking so promising: "Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC).

From Bloomberg, May 8:

• Ongoing discussions with SoftBank have become more advanced
• British semiconductor startup was once valued at $2.8 billion

SoftBank Group Corp. is in talks to acquire Graphcore Ltd., a struggling British semiconductor startup once valued at $2.8 billion, according to people familiar with the deals.

The two companies have held discussions over several months but entered into more advanced deal talks recently, said the people, who asked not to be identified discussing private matters. Financial terms haven’t yet been decided and the talks could still unravel, they said. A final agreement isn’t imminent, one person said.

The talks come amid a surge in sales for SoftBank, thanks largely to its majority stake in another UK-based chip designer, Arm Holdings Plc. On Feb. 7, Arm reported a strong outlook for its expansion beyond smartphones into more artificial intelligence applications. Arm shares have soared by about 40% since.

SoftBank didn’t immediately respond to requests for comment. A representative for Graphcore declined to comment.

Graphcore works on a different type of chip technology than Arm. Formed in Bristol in 2016, the company develops designs for large “intelligence processing units,” meant to help with AI software processing inside data centers. The startup touted its product as a rival to Nvidia Corp.’s high-end graphics chips, and secured high-profile investors including Samsung Electronics Co., Bosch and Sequoia Capital. A 2020 financing round valued Graphcore at $2.8 billion....

....MORE

"IPO Watchlist: European unicorns most likely to list" (incl. Northvolt; Graphcore)

From PitchBook, February 16:

Europe had its worst year for VC-backed IPOs in over a decade due to volatility in the public markets and poor performance from recent listings.

Only 28 companies listed on a stock exchange, raising a total of €1.4 billion ($1.5 billion), according to PitchBook’s 2023 Annual European Venture Report. But as inflation and interest rates come down and the public markets start to stabilize, more VC-backed companies may be getting their ducks in a row with an eye to listing in the near future.

Here are the European VC-backed unicorns most likely to go public. The list was created using PitchBook’s VC Exit Predictor, which calculates exit probability using a machine learning model that is fed historical and real-time data on private company exits. The predictor draws from 34 inputs related to company characteristics, financing and investors....

....MUCH MORE

We've been keeping an eye on Graphcore for eight years. Here's a 2019 post that refers back to that simpler era:

Graphcore CEO Takes A Shot at NVIDIA, Talks His Book, Makes a Good Point (NVDA)

In November 2016 we headlined a post Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA).
Bristol?...

Graphcore was the "Office in Bristol".
A year later: "Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC):
Huh.
Sometimes you get lucky...

They (and Intel) still haven't surpassed NVIDIA but are nipping at the big dog's heels.

And Northvolt? A lot of posts.

Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA)

Bristol?
From HPCwire:

November 1, 2016

Graphcore emerged from stealth mode today with news of a $30 million Series A round to help finance ongoing development of its machine learning (ML) and deep learning acceleration solutions, including a PCIe card that plugs directly into a server’s bus.

The company says the combination of its development framework, called Poplar, and its PCIe-based Intelligent Processing Unit (IPU) can speed up ML and deep learning workloads by 10x to 100x.

The IPU card plugs into the PCI buses of standard X86 servers to provide a processing boost. Armed with multiple IPU cards, a company could enjoy the benefits of “massively parallel, low-precision floating-point compute” at “much higher compute densities” than other solutions.

Graphcore is positioning its IPU cards to take on the workloads that some are looking to run on more exotic hardware, such as graphics processing units (GPUs) or field programmable gate arrays (FPGAs). GPUs and FPGAs represent “stopgap measures” to solving the compute challenges posed by emerging machine intelligence applications, says Graphcore CEO Nigel Toon.

“Our IPU system provides a less restrictive, more efficient, and more powerful solution, making it easier and faster to produce applications, devices and machines that are much more intelligent and which can become more and more useful over time,” Toon says in a press release.

The main problem with GPUs is that they aren’t data-driven. “GPUs have been built to run programs that completely describe the algorithm,” Toon told CNBC. “Machine learning is different. You are trying to teach the system using data and that requires a different style of compute.”

The company has plans to sell two pieces of hardware, including the PCIe-based IPU-Accelerator, as well as an IPU-Alliance that will focus on increasing the performance of the training and inference components of machine intelligence workloads. The company says it will start shipping the appliance next year....MORE

I'm not saying Graphcore is the disruptor, just that it will be something like this. We'll be back with more next week but in the meantime the immediate threat is a bunch of fund managers looking for a bonus wanting to lock in the 280%+ that the stock has made this year and the public exhibiting the round number effect we mentioned back in 2007 as First Solar was on its way from $20 to $317:

...So if you bought when we posted Lazard's upgrade (the stock was $78 and change on its way to $74 before turning) you are on your own.

Gomez Addams ("It's not for nothing that they called me 'the Plunger,'" he boasts, although it was not from Wall Street that he received this nickname, but from the Plumbers Union) was wrong when he'd tell his broker "sell when it gets to $100". FSLR closed yesterday at $99.07 , it's $95 today, the price the insiders and sold at in the recent offering...

NVDA $93.16 down 20 cents.

"A look at Europe's 5 most valuable AI startups"

From PitchBook:

Artificial intelligence & machine learning is perhaps one of the most compelling, disruptive and terrifying technology trends today. It also represents one of the fastest-growing verticals for European venture capital investment. As of August 20, there were 323 VC deals in Europe involving AI & machine-learning businesses, which brought in some €1.9 billion (around $2.1 billion) total. This already exceeds the €1.8 billion raised across 520 deals in the sector last year, which in turn topped 2017's €1.3 billion by a considerable margin, per PitchBook data.

A broad array of startups spanning the value chain are finding their niche in the production and application of AI technology across the vertical. This diversity is witnessed in the top five most valuable European AI startups, which—apart from the fact that they all exist within the same vertical—develop distinct products.

Graphcore (Valuation: $1.7 billion)
Unlike the other companies on this list, Graphcore is primarily concerned with the hardware behind AI tech. The Bristol, England-based startup develops a next generation computer processor to accelerate machine intelligence learning. Given the projected future demand for such processors in this relatively nascent industry, it is little wonder Graphcore holds the higest valuation of AI startups in Europe. It secured unicorn status in December with a $200 million Series D co-led by Atomico and Sofina....

....MUCH MORE

If interested see also:
"Top-10 Artificial Intelligence Startups in Germany"

Top 10 British Artificial Intelligence Startups

The next big thing: VC investing in Central and Eastern Europe

"The Top-10 French Artificial Intelligence Startups"

Climateer Line of the Day: "2018: AI will start separating the winners from the losers" Edition

Via TechCrunch:

“New breakthroughs in AI, enabled by new hardware architectures, will create new intelligent business models for enterprises,” says Nigel Toon, co-founder and CEO at U.K.-based Graphcore.

“Companies that can build an initial knowledge model and launch an initial intelligent service or product, then use this first product to capture new data and improve the knowledge model on a continuing basis, will quickly create clear class-leading products and services that competitors will struggle to keep up with.”

—Enterprise AI will make the leap — who will reap the benefits?, April 17

In November 2016 we headlined a post Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA).
Bristol?...

Graphcore was the "Office in Bristol".
A year later: "Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC):
Huh.
Sometimes you get lucky...

This past January the "Lab in Shenzhen":
AI: "Google moves into Shenzhen in latest China expansion" (GOOG; NVDA)
Not saying that Google's Tensor Processing Unit chips are a threat to NVIDIA (yet) but, at the same time NVIDIA's Mr. Huang was publicly stating the GOOG was no threat, his research peeps were saying "we should make those thingies."

Still waiting on the "Basement in Moscow." 

Chips: AI Deep Learning and More, The New Chip Bestiary

From DataCenterDynamics, January 24:

With AI workloads set to dominate the future, there's an uncertainty around the hardware that’s aiming to dethrone the GPU

In 1971, Intel, then a manufacturer of random access memory, officially released the 4004, its first single-chip central processing unit, thus kickstarting nearly 50 years of CPU dominance in computing.

In 1989, while working at CERN, Tim Berners-Lee used a NeXT computer, designed around the Motorola 68030 CPU, to launch the first website, making the machine used the world’s first web server.

CPUs were the most expensive, the most scientifically advanced, and the most power-hungry parts of a typical server: they became the beating hearts of the digital age, and semiconductors turned into the benchmark for our species' advancement.

This feature appeared in the January issue of DCD Magazine. Subscribe for free today.

Intel's domination

Few might know about the Shannon limit or Landauer's principle, but everyone knows about the existence of Moore’s Law, even if they have never seated a processor in their life. CPUs have entered popular culture and, today, Intel rules this market, with a near-monopoly supported by its massive R&D budgets and extensive fabrication facilities, better known as ‘fabs.’
But in the past two or three years, something strange has been happening: data centers started housing more and more processors that weren’t CPUs.

It began with the arrival of GPUs. It turned out that these massively parallel processors weren’t just useful for rendering video games and mining magical coins, but also for training machines to learn - and chipmakers grabbed onto this new revenue stream for dear life.

Back in August, Nvidia’s CEO Jen-Hsun ‘Jensen’ Huang called AI technologies the “single most powerful force of our time.” During the earnings call, he noted that there were currently more than 4,000 AI start-ups around the world. He also touted examples of enterprise apps that could take weeks to run on CPUs, but just hours on GPUs.

A handful of silicon designers looked at the success of GPUs as they were flying off the shelves, and thought: we can do better. Like Xilinx, a venerable specialist in programming logic devices. The granddaddy of custom silicon, it is credited with inventing the first field-programmable gate arrays (FPGAs) back in 1985.

Applications for FPGAs range from telecoms to medical imaging, hardware emulation, and of course, machine learning workloads. But Xilinx wasn’t happy with adopting old chips for new use cases, the way Nvidia had done, and in 2018, it announced the adaptive compute acceleration platform (ACAP) - a brand new chip architecture designed specifically for AI.

“Data centers are one of several markets being disrupted,” CEO Victor Peng said in a keynote at the recent Xilinx Developer Forum in Amsterdam. “We all hear about the fact that there's zettabytes of data being generated every single month, most of them unstructured. And it takes a tremendous amount of compute capability to process all that data. And on the other side of things, you have challenges like the end of Moore's Law, and power being a problem.

"Because of all these reasons, John Hennessy and Dave Patterson - two icons in the computer science world - both recently stated that we were entering a new golden age of architectural development."
He continued: “Simply put, the traditional architecture that’s been carrying the industry for the last 40 to 50 years is totally inadequate for the level of data generation and data processing that’s needed today.”

“It is important to remember that it’s really, really early in AI,” Peng later told DCD. “There’s a growing feeling that convolutional and deep neural networks aren’t the right approach. This whole black box thing - where you don’t know what’s going on and you can get wildly wrong results, is a little disconcerting for folks.”

A new approach

Salil Raje, head of the Xilinx data center group, warned: “If you’re betting on old hardware and software, you are going to have wasted cycles. You want to use our adaptability and map your requirements to it right now, and then longevity. When you’re doing ASICs, you’re making a big bet.”
Another company making waves is British chip designer Graphcore, quickly becoming one of the most exciting hardware start-ups of the moment.

Graphcore’s GC2 IPU has the world’s highest transistor count for a device that’s actually shipping to customers - 23,600,000,000 of them. That’s not nearly enough to keep up with the demands of Moore’s Law - but it’s a whole lot more transistor gates than in Nvidia’s V100 GPU, or AMD’s monstrous 32-core Epyc CPU.

“The honest truth is, people don’t know what sort of hardware they are going to need for AI in the near future,” Nigel Toon, the CEO of Graphcore, told us in August. “It’s not like building chips for a mature technology challenge. If you know the challenge, you just have to engineer better than other people.

“The workload is very different, neural networks and other structures of interest change from year to year. That’s why we have a research group, it’s sort of a long-distance radar.

"There are several massive technology shifts. One is AI as a workload - we’re not writing programs to tell a machine what to do anymore, we’re writing programs that tell a machine how to learn, and then the machine learns from data. So your programming has gone kind of ‘meta.’ We’re even having arguments across the industry about the way to represent numbers in computers. That hasn’t happened since 1980....

....MUCH MORE 

Top 10 British Artificial Intelligence Startups

The original has the rather pedantic headline "Top-10 British Artificial Intelligence Startups in the UK" and I'm like "What you got against Northern Ireland?" but whatevs.
From Nanalyze:

Once upon a time, the British empire ruled about a quarter of the world’s population. There was a saying: “The sun never sets on the British empire.” While those days of colonialism and conquest are (mostly) behind it, the United Kingdom remains a leading world power, mostly in tea and crumpet consumption. Though, as we understand it, the country still seems to be doing OK financially, boasting the world’s fifth-largest economy. You must admit that the Brits have a certain pragmatic pluck that has always made them competitive for an island nation with limited resources. Now we learn the United Kingdom is the latest country to join the artificial intelligence arms race.

Company	Total Funding	Last Funding Date	Last Funding Amount
OakNorth	$448,500,000	Nov 3, 2017	$125,500,000
BenevolentAI	$202,000,000	Apr 19, 2018	$115,000,000
Darktrace	$179,500,000	Jul 11, 2017	$75,000,000
Graphcore	$110,000,000	Nov 12, 2017	$50,000,000
Blippar	$99,000,000	Mar 2, 2016	$54,000,000
Babylon Health	$85,000,000	Apr 25, 2017	$60,000,000
Qubit	$74,850,000	Feb 22, 2016	$40,000,000
XMOS	$72,390,000	Sep 7, 2017	$15,000,000
Onfido	$60,300,000	Sep 27, 2017	$30,000,000
Callsign	$38,750,000	Jul 27, 2017	$35,000,000

We recently highlighted the top 10 French artificial intelligence startups after that country’s president—who we hear is a big fan of the movie Harold and Maude—announced France would invest $1.8 billion into AI over the next four years. Meanwhile, the Chinese are going all-in on AI, with companies ramping up on hardware like AI chips and other technologies such as computer vision to keep an eye on its nearly 1.4 billion citizens. To keep up with the Jacqueses and the Chans, the Brits have put together a multi-year $1.3 billion package of corporate and venture capital investments to boost the United Kingdom’s AI-based technologies. One of the country’s premiere AI startups was DeepMind before Google acquired the company for about $600 million in 2014, so the Brits have some homegrown talent.We tooled around in Crunchbase a bit, and came up with a list of 10 British AI startups that have taken in the most funding so far.

Lending with AI

 

Topping the list of the top-10 British artificial intelligence startups is one of the UK’s newest unicorns, OakNorth. The three-year-old fintech company has developed an AI-powered platform called ACORN for making data-driven loan decisions for small- to medium-sized businesses. The company has raised a staggering $448.5 million, including a $125 million Series B last November, giving it a valuation of about $1.4 billion. OakNorth even reportedly turned a modest profit of about $14.6 million last year with about $1.65 billion in loans, about quadruple of where it was a year ago. The company recently said it expects to add up to another $2 billion in loans this year and claims it has stimulated the UK economy through its loans by helping create 4,000 jobs and nearly $5.5 billion in economic output.

Its flagship product, ACORN machine, works by pulling in hundreds of data points on whatever industry the loan is to be applied in to analyze the credit risk, which sounds like what other AI fintech startups are doing. That helps make the loan officer become an instant expert in the sector. In addition, the platform constantly monitors other loans being made in the same sector, providing a benchmark and early warning system to proactively monitor risk. OakNorth is also licensing ACORN to other banking institutions, providing yet another potentially huge revenue stream.

Drug Discovery with AI
 
Another British unicorn, with a valuation of about $2 billion after a $115 million funding round in April, BenevolentAI has raised a total of $202 million. That makes the London-based company, founded in 2013, the most valuable private AI drug discovery company in the world. Its algorithms, able to draw upon more than 50 billion contextualised medical facts, can generate insights into the cause of many hard-to-treat diseases. The company claims its platform can cut early stage drug discovery by four years. In addition, it can work more efficiently over the entire drug development process by as much as 60 percent against the pharmaceutical industry average. As you might imagine, that’s not an insignificant cost: A new report published by the Tufts Center for the Study of Drug Development pegs the cost of developing a prescription drug for market approval at $2.6 billion.

Credit: BenevolentAI

The company is targeting everything from Parkinson’s disease to blindness from age-related macular degeneration. Earlier this year, BenevolentAI also acquired a drug discovery and development facility in Cambridge that will allow the startup to begin clinical trial work.

Cybersecurity with AI
 
We did a deep dive into Darktrace about a year ago, and shortly after, the Cambridge-based cybersecurity startup raised another $75 million in a Series D, to bring total funding to $179.5 million, with investors like SoftBank in its corner. Founded in 2013 by some of the country’s bright minds at the University of Cambridge and intelligence experts from spy agencies like the Bond-esque MI5, Darktrace appears to be doing pretty well since we last checked on it, doubling its business under contract and adding more than 200 employees.

Credit: Darktrace

Darktrace takes its inspiration from the human immune system. Our immune system works by learning about what is normal for the body, and then identifies and zaps anything that does not fit the ever-evolving pattern of what qualifies as status quo. Darktrace applies the same logic by employing machine learning and AI algorithms to understand what is normal for every device and user on a network. Anything that does not fit the pattern of normality gets zapped.

Building Chips for AI

 
We first learned about Graphcore, an AI chipmaker out of Bristol, about a year ago for our article on startups building AI hardware. The company, found in 2016, has since taken in another $80 million to bring total funding to $110 million. It also added top VC firm Sequoia Capital to its already impressive list of investors. Graphcore is competing in an increasingly competitive AI processor sector, but seems to have some very promising hardware with its IPU (intelligence processing unit), which it claims can improve performance of machine learning tasks by 10x to 100x compared to others on the market. One test reportedly showed that eight IPU PCIe cards could run a training model in the same amount of time as 128 GPU cards. Watch out, Nvidia (NASDAQ:NVDA)....

...MUCH MORE

"Can Chinese AI Chip Makers Compete with Nvidia?" (NVDA)

Not yet.

However...the fact China not only built the world's fastest supercomputer but did it with chips they designed and manufactured themselves, see 2016's "Milestone: China Builds The (NEW) World's Fastest Supercomuter Using Only Chinese Components (and other news) INTC; NVDA; IBM" combined with our first hit of the three cities named in: November 21, 2016 "Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA)", albeit a year later:

"Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC)

November 13, 2017
BRISTOL, England, Nov. 13, 2017 — Graphcore has today announced a $50 million Series C funding round by Sequoia Capital as the machine intelligence company prepares to ship its first Intelligence Processing Unit (IPU) products to early access customers at the start of 2018....

makes one think the lab in Shenzhen idea is not as far out as it had been.

From Nanalyze, January 11:

There is a new arms race, but we’re not talking thermonuclear war—unless we give machines the launch codes. Artificial intelligence is one of the key technologies that we cover, and it’s been a wild ride the last few years. Industries from healthcare to recruitment have embraced AI to gain efficiencies and a competitive edge. Heck, even Coca Cola is giving you Coke with AI. While the software can be sexy, it’s the hardware, or computing power, which has made many of the advancements in AI possible. One of the companies leading the charge is Nvidia (NASDAQ:NVDA), which has become the gold standard for big computing applications ranging from gaming to supercomputers to neural networks. More than a few AI chip startups have emerged in the last year or two, but the real competition will likely come from established players like AMD or Google. And then there’s China.

Most companies would want to be Nvidia, whose stock gained more than 100 percent in the last year.

We recently told you about all the ways China is Kicking America’s Ass in Tech. We didn’t mention AI because China isn’t there. Yet. However, the Chinese government has plans to reach parity with the United States in AI as early as 2020. It’s putting up the money to do it, starting with a $2 billion AI business park that will be home to 400 enterprises. The country hopes to generate about $60 billion from AI technology by 2025.

Credit: 2017 China-US AI Venture Capital State and Trends Research Report

Despite the official line, Chinese companies are still going with Nvidia, which signed a deal last year to provide its new Volta GPU chips into data centers run by three of China’s biggest tech companies—Alibaba, Baidu and Tencent. It’s little wonder that Nvidia’s stock gained more than 100 percent last year and is off to a scorching start in 2018. Wired reported that Nvidia is in the Chinese government’s crosshairs, urging its industry to develop a chip that is 20 times more powerful and energy efficient than Nvidia’s M40 chip used for artificial neural networks.

A Cambrian Explosion
That goal landed Beijing-based Cambricon Technologies $100 million in funding last August. Alibaba and Lenovo participated in the Series A, which was led by the Chinese government’s largest state-owned investment holding company. That investment propelled Cambricon, founded only in 2016, into the Unicorn Club of companies valued at $1 billion or more. Cambricon hopes to put its AI hardware into one billion smart devices and corner as much as 30 percent of China AI chip market in three years, China Money Market reported. The company recently appeared on CB Insights AI 100 startups list.

Last year, the company released a ton of new products, including three AI processors that can be used in all sorts of applications, from computer vision to autonomous driving to natural language generation. Cambricon also produced a couple of high performance machine learning chips for servers, one market where China lags behind despite being home to more supercomputers than the United States (many of which sport Nvidia hardware). That’s not surprising, as AI Chinese startups like Cambricon have been mainly focused on chips for mobile devices and wearables. In 2016, for example, it made $15 million in licensing fees for its Cambricon-1A chip from smartphone manufacturers and wearable device makers.

A Really, Really Smart Phone
In fact, it makes sense that much of the Chinese AI chip market has been focused on mobile applications, given the country’s emphasis on mobile technology, from social media and e-commerce on WeChat to eSports gaming. Case in point: Semiconductor manufacturer HiSilicon, owned by Chinese telecommunications giant Huawei Technologies, released a wicked fast Kirin 970 processor that features a traditional CPU, a Nvidia-like GPU and an NPU, for neural processing unit, which handles the AI workload. The NPU particularly excels at image recognition, processing a reported 2,000 images per minute. Here how it compares to less-smart smart phones:...MUCH MORE

One thing not mentioned in the Nanalyze report that is critical to understanding Chinese R&D is the importance of the People's Liberation Army (Navy) and the advantage the closed loop of academia, end-user and PLA contracted-and-government-owned companies conveys:

Military AI: China, Russia and the U.S. are Running Neck-and-Neck in an Arms Race

China might be ahead, tough to tell but that's the way to bet.... 

So, we'll try to stay on top of developments and let you know when we see something out oif Shenzhen.

Or Moscow.

"China overtakes US in AI startup funding with a focus on facial recognition and chips"

NVIDIA watches.
And yours truly writes stuff like Jan. 15's "'Can Chinese AI Chip Makers Compete with Nvidia?' (NVDA)":

Not yet.
However...the fact China not only built the world's fastest supercomputer but did it with chips they designed and manufactured themselves, see 2016's "Milestone: China Builds The (NEW) World's Fastest Supercomuter Using Only Chinese Components (and other news) INTC; NVDA; IBM" combined with our first hit of the three cities named in: November 21, 2016 "Artificial Intelligence: What Could Derail NVIDIA? A Lab in Shenzhen; A Basement in Moscow; An Office in Bristol (NVDA)", albeit a year later:

"Sequoia Backs Graphcore as the Future of Artificial Intelligence Processors" (NVDA; INTC)

November 13, 2017
BRISTOL, England, Nov. 13, 2017 — Graphcore has today announced a $50 million Series C funding round by Sequoia Capital as the machine intelligence company prepares to ship its first Intelligence Processing Unit (IPU) products to early access customers at the start of 2018....

makes one think the lab in Shenzhen idea is not as far out as it had been....

And today's headliner from The Verge:

The competition between China and the US in AI development is tricky to quantify. While we do have some hard numbers, even they are open to interpretation. The latest comes from technology analysts CB Insights, which reports that China has overtaken the US in the funding of AI startups. The country accounted for 48 percent of the world’s total AI startup funding in 2017, compared to 38 percent for the US. 

It’s not a straightforward victory for China, however. In terms of the volume of individual deals, the country only accounts for 9 percent of the total, while the US leads in both the total number of AI startups and total funding overall. The bottom line is that China is ahead when it comes to the dollar value of AI startup funding, which CB Insights says shows the country is “aggressively executing a thoroughly-designed vision for AI.”

China’s natural advantages in AI are well-documented. Compared to the US, it has a huge population (1.4 billion), which offers a wealth of data and opportunity for companies to scale quickly. Its AI sector also has the backing of a central government that’s able to quickly shift resources (as opposed to the missing-in-action White House), and the country’s looser approach to digital regulations means companies can experiment more freely....MORE 

Now, about that basement in Moscow...

More to come.