NeuReality lands $35M to bring AI accelerator chips to market

The growing demand for AI, particularly generative AI (i.e., AI that generates images, text and more), is supercharging the AI inferencing chip market. Inferencing chips accelerate the AI inferencing process, which is where AI systems generate outputs (e.g., text, images, audio) based on what they learned while “training” on a specific set of data. AI inferencing chips can be — and have been — used to yield faster generations from systems such as Stable Diffusion, which translates text prompts into artwork, and OpenAI’s GPT-3, which extends a few lines of prose into full-length poems, essays and more.

A number of vendors — both startups and well-established players — are actively developing and selling access to AI inferencing chips. There’s Hailo, Mythic and Flex Logix, to name a few upstarts. And on the incumbent side, Google’s competing for dominance with its tensor processing units (TPUs) while Amazon’s betting on Inferentia. But the competition, while fierce, hasn’t scared away firms like NeuReality, which occupy the AI chip inferencing market but aim to differentiate themselves by offering a suite of software and services to support their hardware.

On the subject, NeuReality today announced that it raised $35 million in a Series A funding round led by Samsung Ventures, Cardumen Capital, Varana Capital, OurCrowd and XT Hi-Tech with participation from SK Hynix, Cleveland Avenue, Korean Investment Partners, StoneBridge, and Glory Ventures. Co-founder and CEO Moshe Tanach tells TechCrunch that the tranche will be put toward finalizing the design of NeuReality’s flagship AI inferencing chip in early 2023 and shipping it to customers.

NeuReality was founded with the vision to build a new generation of AI inferencing solutions that are unleashed from traditional CPU-centric architectures and deliver high performance and low latency, with the best possible efficiency in cost and power consumption,” Tanach told TechCrunch via email. “Most companies that can leverage AI don’t have the funds nor the huge R&D that Amazon, Meta and other huge companies investing in AI have. NeuReality will bring AI tech to anyone who wants to deploy easily and affordably.”

NeuReality was co-founded in 2019 by Tzvika Shmueli, Yossi Kasus and Tanach, who previously served as a director of engineering at Marvell and Intel. Shmueli was formerly the VP of back-end infrastructure at Mellanox Technologies and the VP of engineering at Habana Labs. As for Kasus, he held a senior director of engineering role at Mellanox and was the head of integrations at semiconductor company EZchip.

From the start, NeuReality focused on bringing to market AI hardware for cloud data centers and “edge” computers, or machines that run on-premises and do most of their data processing offline. Tanach says that the startup’s current-generation product lineup, the Network Attached Processing Unit (NAPU), is optimized for AI inference applications, including computer vision (think algorithms that recognize objects in photos), natural language processing (text-generating and classifying systems) and recommendation engines (like the type that suggest products on e-commerce sites).

NeuReality’s NAPU is essentially a hybrid of multiple types of processors. It can perform functions like AI inferencing load balancing, job scheduling and queue management, which have traditionally been done in software but not necessarily very efficiently.

NeuReality

Image Credits: NeuReality

NeuReality’s NR1, an FPGA-based SKU within the NAPU family, is a network-attached “server on a chip” with an embedded AI inferencing accelerator along with networking and virtualization capabilities. NeuReality also offers the NR1-M module, a PCIe card containing an NR1 and a network-attached inference server, and a separate module — the NR1-S — that pairs several NR1-Ms with the NR1.

On the software side, NeuReality delivers a set of tools, including a software development kit for cloud and local workloads, a deployment manager to help with runtime issues and a monitoring dashboard.

“The software for AI inference [and] the tools for heterogeneous compute and automated flow of compilation and deployment … is the magic that supports our innovative hardware approach,” Tanach said. “The first beneficiaries of the NAPU technology are enterprises and cloud solution providers that need infrastructure to support their chatbots, voice bots, automatic transcriptions and sentiment analysis as well as computer vision use cases for document scans, defect detection, etc.  … While the world was focusing on the deep learning processor improvements, NeuReality focused on optimizing the system around it and the software layers above it to provide higher efficiency and a much easier flow to deploy inference.”

NeuReality, it must be noted, has yet to back up some of its performance claims with empirical evidence. It told ZDNet in a recent article that it estimates its hardware will deliver a 15x improvement in performance per dollar compared to the available GPUs and ASICs offered by deep learning accelerator vendors, but NeuReality hasn’t released validating benchmarking data. The startup also hasn’t detailed its proprietary networking protocol, a protocol that it has previously claimed is more performant than existing solutions.

Those items aside, delivering hardware at massive scale isn’t easy — particularly where it involves custom AI inferencing chips. But Tanach argues that NeuReality has laid the necessary groundwork, partnering with AMD-owned semiconductor manufacturer Xilinx for production and inking a partnership with IBM to work on hardware requirements for the NR1. (IBM, which is also a NeuReality design partner, previously said it’s “evaluating” the startup’s products for use in the IBM cloud.) NeuReality has been shipping prototypes to partners since May 2021, Tanach says.

According to Tanach, beyond IBM, NeuReality is working with Lenovo, AMD and unnamed cloud solution providers, system integrators, deep learning accelerator vendors and “inference-consuming” enterprises on deployments. Tanach declined, however, to reveal how many customers the startup currently has or what roughly it’s projecting in terms of revenue.

“We see that the pandemic is slowing companies down and pushing for consolidation between the many deep learning vendors. However, for us it doesn’t change anything, since late next year or sometime through 2024 inference deployment is expected to explode — and our technology is exactly the enabler and driver of that growth,” Tanach said. “The NAPU will bring AI for a broader set of less technical companies. It is also set to allow large-scale users such as ‘hyperscalers’ and next-wave data center customers to support their growing scale of AI usage.”

Ori Kirshner, the head of Samsung Ventures in Israel, added in an emailed statement: “We see substantial and immediate need for higher efficiency and easy-to-deploy inference solutions for data centers and on-premises use cases, and this is why we are investing in NeuReality. The company’s innovative disaggregation, data movement and processing technologies improve computation flows, compute-storage flows, and in-storage compute — all of which are critical for the ability to adopt and grow AI solutions.”

NeuReality, which currently has 40 employees, plans to hire 20 more over the next two fiscal quarters. To date, it’s raised $38 million in venture capital.

NeuReality lands $35M to bring AI accelerator chips to market by Kyle Wiggers originally published on TechCrunch

Pliops lands $100M for chips that accelerate analytics in data centers

Analyzing data generated within the enterprise — for example, sales and purchasing data — can lead to insights that improve operations. But some organizations are struggling to process, store and use their vast amounts of data efficiently. According to an IDC survey commissioned by Seagate, organizations collect only 56% of the data available throughout their lines of business, and out of that 56%, they only use 57%.

Part of the problem is that data-intensive workloads require substantial resources, and that adding the necessary compute and storage infrastructure is often expensive. For companies moving to the cloud specifically, IDG reports that they plan to devote $78 million toward infrastructure this year. Thirty-six percent cited controlling costs as their top challenge.

That’s why Uri Beitler launched Pliops, a startup developing what he calls “data processors” for enterprise and cloud data centers. Pliop’s processors are engineered to boost the performance of databases and other apps that run on flash memory, saving money in the long run, he claims.

“It became clear that today’s data needs are incompatible with yesterday’s data center architecture. Massive data growth has collided with legacy compute and storage shortcomings, creating slowdowns in computing, storage bottlenecks and diminishing networking efficiency,” Beitler told TechCrunch in an email interview. “While CPU performance is increasing, it’s not keeping up, especially where accelerated performance is critical. Adding more infrastructure often proves to be cost prohibitive and hard to manage. As a result, organizations are looking for solutions that free CPUs from computationally intensive storage tasks.”

Pliops isn’t the first to market with a processor for data analytics. Nvidia sells the BlueField-3 data processing unit (DPU). Marvell has its Octeon technology. Oracle’s SPARC M7 chip has a data analytics accelerator coprocessor with a specialized set of instructions for data transformation. And in the realm of startups, Blueshift Memory and Speedata are creating hardware that they say can perform analytics tasks significantly faster than standard processors.

Pliops

Image Credits: Pliops

But Pliops claims to be further along than most, with deployments and pilots with customers (albeit unnamed) including fintechs, “medium-sized” communication service providers, data center operators and government labs. The startup’s early traction won over investors, it would seem, which poured $100 million into its Series D round that closed today.

Koch Disruptive Technologies led the tranche, with participation from SK Hynix and Walden International’s Lip-Bu Tan, bringing Pliops’ total capital raised to date to more than $200 million. Beitler says that it’ll be put toward building out the company’s hardware and software roadmap, bolstering Pliops’ footprint with partners and expanding its international headcount.

“Many of our customers saw tremendous growth during the COVID-19 pandemic, thanks in part to their ability to react quickly to the new work environment and conditions of uncertainty. Pliops certainly did. While some customers were affected by supply chain issues, we were not,” Beitler said. “We do not see any slowdown in data growth — or the need to leverage it. Pliops was strong before this latest funding round and even stronger now.”

Accelerating data processing

Beitler, the former director of advanced memory solutions at Samsung’s Israel Research Center, co-founded Pliops in 2017 alongside Moshe Twitto and Aryeh Mergi. Twitto was a research scientist at Samsung developing signal processing technologies for flash memory, while Mergi co-launched a number of startups — including two that were acquired by EMC and SanDisk — prior to joining Pliops.

Pliop’s processor delivers drive fail protection for solid-state drives (SSD) as well as in-line compression, a technology that shrinks the size of data by finding identical data sequences and then saving only the first sequence. Beitler claims the company’s technology can reduce drive space while expanding capacity, mapping “variable-sized” compressed objects within storage to reduce wasted space.

A core component of Pliops’ processor is its hardware-accelerated key-value storage engine. In key-value databases — databases where data is stored in a “key-value” format and optimized for reading and writing — key-value engines manage all the persistent data directly. Beitler makes the case that CPUs are typically over-utilized when running these engines, resulting in apps not taking full advantage of SSD’s capabilities.

“Organizations are looking for solutions that free CPUs from computationally-intensive storage tasks. Our hardware helps create a modern data center architecture by leveraging a new generation of hardware-accelerated data processing and storage management technology — one that delivers orders of magnitude improvement in performance, reliability and scalability,” Beitler said. “In short, Pliops enables getting more out of existing infrastructure investments.”

Pliops’ processor became commercially available last July. The development team’s current focus is accelerating the ingest of data for machine learning use cases, Beitler says — use cases that have grown among Pliops’ current and potential customers.

The road ahead

Certainly, Pliops has its work cut out for it. Nvidia is a formidable competitor in the data processing accelerator space, having spent years developing its BlueField lineup. And AMD acquired DPU vendor Pensando for $1.9 billion, signaling its wider ambitions.

A move that could pay dividends for Pliops is joining the Open Programmable Infrastructure Project (OPI), a relatively new venture under the Linux Foundation that aims to create standards around data accelerator hardware. While Pliops isn’t a member yet — current members include Intel, Nvidia, Marvell, F5, Red Hat, Dell and Keysight Technologies — it stands to reason that becoming one could expose its technology to a larger customer base.

Beitler demurred when asked about OPI, but pointed out that the market for data acceleration is still nascent and growing.

“We continue to see both infrastructure and application teams being overwhelmed with underperforming storage and overwhelmed applications that aren’t meeting company’s data demands,” Beitler said. “The overall feedback is that our processor is a game-changing product and without it companies are required to make years of investments in software and hardware engineering to solve the same problem.”

SK Hynix gets China approval to take over Intel’s NAND business 

South Korean chipmaker SK Hynix said on Wednesday it has received merger clearance from the Chinese antitrust authority for its $9 billion acquisition of Intel Corp’s NAND and SSD (solid-state drive) business, clearing the final hurdle to completion of securing regulatory approval in eight jurisdictions. 

Last October, the U.S. chip giant and SK Hynix reached the acquisition deal. Following the agreement, SK Hynix obtained nods from watchdog agencies in South Korea, the U.S., the European Union, Taiwan, Brazil, Britain and Singapore. 

SK Hynix said in its statement: “SK Hynix sincerely welcomes and appreciates the State Administration of Market Regulation’s merger clearance for the deal. SK Hynix will enhance its competitiveness of NAND Flash and SSD business by continuing the remaining post-merger integration process.” 

The acquisition, which is SK Hynix’s largest acquisition deal, will help SK Hynix to expand its NAND SSD business and narrow the gap with market leader Samsung Electronics. Meanwhile, Intel will continue to retain optane business to invest in more advanced technology, Intel said last year. The U.S. company plans to divest the NAND unit and double down on developing technology, including 5G network infrastructure, artificial intelligence and edge computing. 

SK Hynix will make the first payment of $7 billion by the end of this year and the remaining $2 billion by March 2025, a spokesperson at SK Hynix confirmed. Once the deal completes, the South Korean chipmaker will take over Intel’s NAND SSD, NAND component and wafer businesses (including NAND-related intellectual property and employees), and its NAND memory manufacturing facility in Dalian.  

There has been concern that SK Hynix will not receive China’s permission for the deal amid the tensions between the U.S. and China. SK Hynix said the approval comes at the right time without a significant delay as the deal “is deemed mutually beneficial” for all three countries. 

China’s State Administration for Market Regulation said in its statement on Wednesday that it had approved, but with a number of conditions that will last for five years. 

The conditions include that SK Hynix should expand its production quantity of PCIe and SATA enterprise-class solid-state hard disks products, and supply the products at fair, reasonable and nondiscriminatory prices, according to the statement. It also said that SK Hynix should not force customers in China to exclusively purchase products from SK Hynix or companies controlled by SK Hynix.   

Intel agrees to sell its NAND business to SK Hynix for $9 billion

SK Hynix, one of the world’s largest chip makers, announced today it will pay $9 billion for Intel’s flash memory business. Intel said it will use proceeds from the deal to focus on artificial intelligence, 5G and edge computing.

“For Intel, this transaction will allow us to to further prioritize our investments in differentiated technology where we can play a bigger role in the success of our customers and deliver attractive returns to our stockholders,” said Intel chief executive officer Bob Swan in the announcement.

The Wall Street Journal first reported earlier this week that the two companies were nearing an agreement, which will turn SK Hynix into one of the world’s largest NAND memory makers, second only to Samsung Electronics.

The deal with SK Hynix is the latest one Intel has made so it can double down on developing technology for 5G network infrastructure. Last year, Intel sold the majority of its modem business to Apple for about $1 billion, with Swan saying that the time that the deal would allow Intel to “[put] our full effort into 5G where it most closely aligns with the needs of our global customer base.”

Once the deal is approved and closes, Seoul-based SK Hynix will take over Intel’s NAND SSD and NAND component and wafer businesses, and its NAND foundry in Dalian, China. Intel will hold onto its Optane business, which makes SSD memory modules. The companies said regulatory approval is expected by late 2021, and a final closing of all assets, including Intel’s NAND-related intellectual property, will take place in March 2025.

Until the final closing takes places, Intel will continue to manufacture NAND wafers at the Dalian foundry and retain all IP related to the manufacturing and design of its NAND flash wafers.

As the Wall Street Journal noted, the Dalian facility is Intel’s only major foundry in China, which means selling it to SK Hynix will dramatically reduce its presence there as the United States government puts trade restrictions on Chinese technology.

In the announcement, Intel said it plans to use proceeds from the sale to “advance its long-term growth priorities, including artificial intelligence, 5G networking and the intelligent, autonomous edge.”

During the six-month period ending on June 27, 2020, NAND business represented about $2.8 billion of revenue for its Non-volatile Memory Solutions Group (NSG), and contributed about $600 million to the division’s operating income. According to the Wall Street Journal, this made up the majority of Intel’s total memory sales during that period, which was about $3 billion.

SK Hynix CEO Seok-Hee Lee said the deal will allow the South Korean company to “optimize our business structure, expanding our innovative portfolio in the NAND flash market segment, which will be comparable with what we achieved in DRAM.”

Intel agrees to sell its NAND business to SK Hynix for $9 billion

SK Hynix, one of the world’s largest chip makers, announced today it will pay $9 billion for Intel’s flash memory business. Intel said it will use proceeds from the deal to focus on artificial intelligence, 5G and edge computing.

“For Intel, this transaction will allow us to to further prioritize our investments in differentiated technology where we can play a bigger role in the success of our customers and deliver attractive returns to our stockholders,” said Intel chief executive officer Bob Swan in the announcement.

The Wall Street Journal first reported earlier this week that the two companies were nearing an agreement, which will turn SK Hynix into one of the world’s largest NAND memory makers, second only to Samsung Electronics.

The deal with SK Hynix is the latest one Intel has made so it can double down on developing technology for 5G network infrastructure. Last year, Intel sold the majority of its modem business to Apple for about $1 billion, with Swan saying that the time that the deal would allow Intel to “[put] our full effort into 5G where it most closely aligns with the needs of our global customer base.”

Once the deal is approved and closes, Seoul-based SK Hynix will take over Intel’s NAND SSD and NAND component and wafer businesses, and its NAND foundry in Dalian, China. Intel will hold onto its Optane business, which makes SSD memory modules. The companies said regulatory approval is expected by late 2021, and a final closing of all assets, including Intel’s NAND-related intellectual property, will take place in March 2025.

Until the final closing takes places, Intel will continue to manufacture NAND wafers at the Dalian foundry and retain all IP related to the manufacturing and design of its NAND flash wafers.

As the Wall Street Journal noted, the Dalian facility is Intel’s only major foundry in China, which means selling it to SK Hynix will dramatically reduce its presence there as the United States government puts trade restrictions on Chinese technology.

In the announcement, Intel said it plans to use proceeds from the sale to “advance its long-term growth priorities, including artificial intelligence, 5G networking and the intelligent, autonomous edge.”

During the six-month period ending on June 27, 2020, NAND business represented about $2.8 billion of revenue for its Non-volatile Memory Solutions Group (NSG), and contributed about $600 million to the division’s operating income. According to the Wall Street Journal, this made up the majority of Intel’s total memory sales during that period, which was about $3 billion.

SK Hynix CEO Seok-Hee Lee said the deal will allow the South Korean company to “optimize our business structure, expanding our innovative portfolio in the NAND flash market segment, which will be comparable with what we achieved in DRAM.”

W(hy)TF are Japan and South Korea in a trade war?

Another week, another trade war. And unlike most trade wars these days, this one didn’t originate from the confines of the Rose Garden with the Marine One whirlybird in the background. No, like any Ice Bucket Challenge-worthy meme, others are getting in on the trade war bandwagon and making it their own.

Cue Japan and South Korea. The two countries have slipped into their own trade war over the past few weeks, a conflict that now threatens the foundations of Japan’s supplier industry, Samsung Electronics, and global smartphone and computer shipments.

But why a trade conflict? If the U.S./China trade war emanates from the dark recesses of President Trump’s brain, then this new trade war emanates from the dark chapters of Japan and South Korea’s collective and sad history.

One of the saddest of those chapters is the plight of Korean comfort women — women who were forced into sexual slavery by wartime Japan in the 1930s and 1940s to service soldiers throughout the Japanese empire. Given the dates of those atrocities, many of those women are now reaching the late stages of their lives, as are men who were impressed into wartime labor in Japanese factories to fight the Allies.

Late last year, Korea’s highest court ordered Mitsubishi to pay essentially reparations for the company’s use of slave labor throughout the Japanese occupation and World War II, a decision that mirrored the court’s earlier judgment against Nippon Steel & Sumitomo Metal a few weeks before.

As the Korean court system has attempted to claw back those reparations from Japanese companies, Japan has not sat still. The country’s prime minister Shinzo Abe and his government have responded by placing a broad trade embargo on South Korea of high-technology goods under “national security” grounds, arguing that Seoul has failed to find a path forward to mend the fences between the two countries.

This past week, the two countries met to try to resolve the tensions, but failed to agree on a solution. That leaves the export bans in place, jeopardizing the supply chains for many electronics products.

Take Samsung Electronics for instance. The Korean company is the number one manufacturer of memory DRAM chips, accounting for more than 40% of the nearly $100 billion market, and also the number one manufacturer of NAND flash chips, with 35% share. SK Hynix — another Korean company — was the second largest manufacturer of DRAM chips with a roughly 31% share. Samsung and other Korean manufacturers are also market leading in industries like semiconductors and LCD displays.

Korea’s electronics companies have deep supply chains in Japan, which produce everything from photoresist chemicals and materials for semiconductors to the actual manufacturing equipment and parts required to operate factories. Thus, Japan’s trade embargo was expected to compromise two of Korea’s leading manufacturers, a punch to Korea’s fragile economy and a wake-up call for President Moon to reach a compromise with Prime Minister Abe.

Except, as often happens in the wacky world of trade, the export ban had unexpectedly positive consequences.

An anticipated glut of DRAM memory chips this year had pushed prices to new lows, slashing profits at Samsung Electronics in the company’s worst drop in four years. The company’s stock has been battered: from August last year until January, the company lost a third of its value.

And then Japan interceded. Supplies of DRAM chips are suddenly dropping — and prices are rising in turn. As the Wall Street Journal noted Thursday, Japan’s curbs are actually shoring up the memory chip market and leading to better than expected results for Samsung and other Korean manufacturers. While it has had a topsy-turvy few weeks, the stock price for Samsung Electronics is now almost back to where it was this time last year.

In other words, Japan’s punch was more like a stimulus. Whoops.

Such short-term gains may be amusing for trade policy watchers, but any returns are likely to be short-lived of course. And the news is much worse for semiconductors. As the Nikkei Asian Review noted this week, “Any disruption in the supply of EUV photoresist — a coating product used in the extreme ultraviolet lithography vital to the most complex semiconductors — could set back Samsung’s plans to launch its 7-nanometer chips around the turn of the year.” The company has stockpiled some materials, but if the trade war extends from weeks to months, it will eventually have to succumb from the damage to its supply chain.

All of which is to say that what started as a trade spat might boil over into shrinking quantities of memory chips, displays, and next-generation semiconductors — in other words, pretty much everything you need to build a computer or smartphone today.

There are a couple of lessons for the tech industry here. First, while Silicon Valley and other tech regions enjoy a mostly ahistorical outlook, the antecedents of the world are always brimming just beneath the surface. The comfort women situation may seem tangential to the day-to-day challenges of building a hardware product, but politics — particularly visceral, human politics — has a way of interceding far from its remit.

Second, even in a globalized world where national politicians lust for economic growth (and certainly Prime Minister Abe and President Moon are heavily invested in growing their respective economies), networked and cross-border supply chains are increasingly fragile. Just as Huawei discovered the dangers of relying on American technology over the past year, now Korean companies are learning about the dangers of depending on Japan’s high technology industry for critical components.

Third, the development of 5G wireless technology standards and associated hardware devices just increasingly gets battered. The U.S. has specifically targeted Huawei over 5G, but Samsung also has 5G modems and network equipment underway, which are now threatened in Japan and South Korea’s trade war. As wireless technology has become essential to global commerce and entertainment the past few decades, the political importance of controlling this technology has increased dramatically.

Ultimately, what’s the resolution to this new trade war? Well, that’s part of the challenge. President Moon doesn’t want to agree to a quick truce, worrying that such a rapid negotiation would appear to be giving in to Japan’s demands — a symbolism that he is unlikely to accept. Meanwhile, Prime Minister Abe faces the opposite forces, with the Japanese government holding the line that all claims to reparations over the comfort women and wartime slavery were settled by the two countries’ bilateral trade agreement from the 1960s and other diplomatic agreements.

Yet, both politicians need economic growth to succeed, and compromising their leading companies from selling their leading exports is not a route to that outcome. Both are principled leaders, but both are ultimately pragmatic. And so as it happens, it may not be the State Department that gets a deal over the line. No, maybe it’s time Tim Cook gets on his iPhone and talks about, well, iPhones.

W(hy)TF are Japan and South Korea in a trade war?

Another week, another trade war. And unlike most trade wars these days, this one didn’t originate from the confines of the Rose Garden with the Marine One whirlybird in the background. No, like any Ice Bucket Challenge-worthy meme, others are getting in on the trade war bandwagon and making it their own.

Cue Japan and South Korea. The two countries have slipped into their own trade war over the past few weeks, a conflict that now threatens the foundations of Japan’s supplier industry, Samsung Electronics, and global smartphone and computer shipments.

But why a trade conflict? If the U.S./China trade war emanates from the dark recesses of President Trump’s brain, then this new trade war emanates from the dark chapters of Japan and South Korea’s collective and sad history.

One of the saddest of those chapters is the plight of Korean comfort women — women who were forced into sexual slavery by wartime Japan in the 1930s and 1940s to service soldiers throughout the Japanese empire. Given the dates of those atrocities, many of those women are now reaching the late stages of their lives, as are men who were impressed into wartime labor in Japanese factories to fight the Allies.

Late last year, Korea’s highest court ordered Mitsubishi to pay essentially reparations for the company’s use of slave labor throughout the Japanese occupation and World War II, a decision that mirrored the court’s earlier judgment against Nippon Steel & Sumitomo Metal a few weeks before.

As the Korean court system has attempted to claw back those reparations from Japanese companies, Japan has not sat still. The country’s prime minister Shinzo Abe and his government have responded by placing a broad trade embargo on South Korea of high-technology goods under “national security” grounds, arguing that Seoul has failed to find a path forward to mend the fences between the two countries.

This past week, the two countries met to try to resolve the tensions, but failed to agree on a solution. That leaves the export bans in place, jeopardizing the supply chains for many electronics products.

Take Samsung Electronics for instance. The Korean company is the number one manufacturer of memory DRAM chips, accounting for more than 40% of the nearly $100 billion market, and also the number one manufacturer of NAND flash chips, with 35% share. SK Hynix — another Korean company — was the second largest manufacturer of DRAM chips with a roughly 31% share. Samsung and other Korean manufacturers are also market leading in industries like semiconductors and LCD displays.

Korea’s electronics companies have deep supply chains in Japan, which produce everything from photoresist chemicals and materials for semiconductors to the actual manufacturing equipment and parts required to operate factories. Thus, Japan’s trade embargo was expected to compromise two of Korea’s leading manufacturers, a punch to Korea’s fragile economy and a wake-up call for President Moon to reach a compromise with Prime Minister Abe.

Except, as often happens in the wacky world of trade, the export ban had unexpectedly positive consequences.

An anticipated glut of DRAM memory chips this year had pushed prices to new lows, slashing profits at Samsung Electronics in the company’s worst drop in four years. The company’s stock has been battered: from August last year until January, the company lost a third of its value.

And then Japan interceded. Supplies of DRAM chips are suddenly dropping — and prices are rising in turn. As the Wall Street Journal noted Thursday, Japan’s curbs are actually shoring up the memory chip market and leading to better than expected results for Samsung and other Korean manufacturers. While it has had a topsy-turvy few weeks, the stock price for Samsung Electronics is now almost back to where it was this time last year.

In other words, Japan’s punch was more like a stimulus. Whoops.

Such short-term gains may be amusing for trade policy watchers, but any returns are likely to be short-lived of course. And the news is much worse for semiconductors. As the Nikkei Asian Review noted this week, “Any disruption in the supply of EUV photoresist — a coating product used in the extreme ultraviolet lithography vital to the most complex semiconductors — could set back Samsung’s plans to launch its 7-nanometer chips around the turn of the year.” The company has stockpiled some materials, but if the trade war extends from weeks to months, it will eventually have to succumb from the damage to its supply chain.

All of which is to say that what started as a trade spat might boil over into shrinking quantities of memory chips, displays, and next-generation semiconductors — in other words, pretty much everything you need to build a computer or smartphone today.

There are a couple of lessons for the tech industry here. First, while Silicon Valley and other tech regions enjoy a mostly ahistorical outlook, the antecedents of the world are always brimming just beneath the surface. The comfort women situation may seem tangential to the day-to-day challenges of building a hardware product, but politics — particularly visceral, human politics — has a way of interceding far from its remit.

Second, even in a globalized world where national politicians lust for economic growth (and certainly Prime Minister Abe and President Moon are heavily invested in growing their respective economies), networked and cross-border supply chains are increasingly fragile. Just as Huawei discovered the dangers of relying on American technology over the past year, now Korean companies are learning about the dangers of depending on Japan’s high technology industry for critical components.

Third, the development of 5G wireless technology standards and associated hardware devices just increasingly gets battered. The U.S. has specifically targeted Huawei over 5G, but Samsung also has 5G modems and network equipment underway, which are now threatened in Japan and South Korea’s trade war. As wireless technology has become essential to global commerce and entertainment the past few decades, the political importance of controlling this technology has increased dramatically.

Ultimately, what’s the resolution to this new trade war? Well, that’s part of the challenge. President Moon doesn’t want to agree to a quick truce, worrying that such a rapid negotiation would appear to be giving in to Japan’s demands — a symbolism that he is unlikely to accept. Meanwhile, Prime Minister Abe faces the opposite forces, with the Japanese government holding the line that all claims to reparations over the comfort women and wartime slavery were settled by the two countries’ bilateral trade agreement from the 1960s and other diplomatic agreements.

Yet, both politicians need economic growth to succeed, and compromising their leading companies from selling their leading exports is not a route to that outcome. Both are principled leaders, but both are ultimately pragmatic. And so as it happens, it may not be the State Department that gets a deal over the line. No, maybe it’s time Tim Cook gets on his iPhone and talks about, well, iPhones.

Korean conglomerate SK leads $600M round for Chinese chipmaker Horizon Robotics

Horizon Robotics, a three-year-old Chinese startup backed by Intel Capital, just raised a mega-round of fundings from domestic and overseas backers as it competes for global supremacy in developing AI solutions and chips aimed at autonomous vehicles, smart retail stores, surveillance equipment and other devices for everyday scenarios.

The Beijing-based company announced Wednesday in a statement that it’s hauled in $600 million in a Series B funding round led by SK China, the China subsidiary of South Korean conglomerate SK Group; SK Hynix, SK’s semiconductor unit; and a number of undisclosed Chinese automakers along with their funds.

The fresh capital drove Horizon’s valuation to at least $3 billion, the company claims. The Financial Times previously reported that the chipmaker was raising up to $1 billion in a funding round that could value it at as much as $4 billion. Such a price tag could perhaps be justified by the vast amount of resources China has poured into the red-hot sector as part of a national push to shed dependency on imported chips and work towards what analysts call “semiconductor sovereignty.”

Horizon did not specify how the proceeds will be used. The company could not be immediately reached for comments.

In 2015, Yu Kai left Baidu as the Chinese search engine giant’s deep learning executive and founded Horizon to make the “brains” for a broad spectrum of connected devices. In doing so Yu essentially set himself up for a race against industry veterans like Intel and Nvidia. To date, the startup has managed to make a dent by securing government contracts, which provide a stable source of income for China’s AI upstarts including SenseTime, and several big-name clients like SK’s telecommunication unit, which is already leveraging Horizon’s algorithms to develop smart retail solutions. Like many of its peers who are at the forefront of the AI race, Horizon has set up an office in Silicon Valley and hiring local talents for its lab.

Other investors who joined the round included several of Horizon’s existing backers such as Hillhouse Capital and Morningside Venture Capital, an investment fund run by Chinese conglomerate China Oceanwide Holdings, and the CSOBOR Fund, a private equity firm set up by China’s state-owned conglomerate CITIC to back projects pertaining to China’s ambitious “One Belt, One Road” modern Silk Road initiative.

Nvidia’s limited China connections

Another round of followups on Nvidia, and then some short news analysis.

TechCrunch is experimenting with new content forms. This is a rough draft of something new – provide your feedback directly to the author (Danny at danny@techcrunch.com) if you like or hate something here.

Nvidia / TSMC questions

Following up on my analyses this week on Nvidia (Part 1, Part 2) , a reader asked in regards to Nvidia’s risk with China tariffs:

but the TSMC impact w.r.t. tariffs doesn’t make sense to me. TSMC is largely not impacted by tariffs and so the supply chain with NVIDIA is also not impacted w.r.t. to TSMC as a supplier. There are many alternate wafer suppliers in Taiwan.

This is a challenging question to definitively answer, since obviously Nvidia doesn’t publicly disclose its supply chain, or more granularly, which factories those supply chain partners utilize for its production. It does, however, list a number of companies in its 10-K form as manufacturing, testing, and packaging partners, including:

To understand how this all fits together, there are essentially three phases for bringing a semiconductor to market:

  1. Design – this is Nvidia’s core specialty
  2. Manufacturing – actually making the chip from silicon and other materials at the precision required for it to be reliable
  3. Testing, packaging and distribution – once chips are made, they need to be tested to prove that manufacturing worked, then packaged properly to protect them and shipped worldwide to wherever they are going to be assembled/integrated

For the highest precision manufacturing required for chips like Nvidia’s, Taiwan, South Korea and the U.S. are the world leaders, with China trying to catch up through programs like Made in China 2025 (which, after caustic pushback from countries around the world, it looks like Beijing is potentially scrapping this week). China is still considered to be one-to-two generations behind in chip manufacturing, though it increasingly owns the low-end of the market.

Where the semiconductor supply chain traditionally gets more entwined with China is around testing and packaging, which are generally considered lower value (albeit critical) tasks that have been increasingly outsourced to the mainland over the years. Taiwan remains the dominant player here as well, with roughly 50% of the global market, but China has been rapidly expanding.

U.S. tariffs on Chinese goods do not apply to Taiwan, and so for the most part, Nvidia’s supply chain should be adept at avoiding most of the brunt of the trade conflict. And while assembly is heavily based in China, electronics assemblers are rapidly adapting their supply chains to mitigate the damage of tariffs by moving factories to Vietnam, India, and elsewhere.

Where it gets tricky is the Chinese market itself, which imports a huge number of semiconductor chips, and represents roughly 20% of Nvidia’s revenues. Even here, many analysts believe that the Chinese will have no choice but to buy Nvidia’s chips, since they are market-leading and substitutes are not easily available.

So the conclusion is that Nvidia likely has maneuvering room in the short-term to weather exogenous trade tariff shocks and mitigate their damage. Medium to long-term though, the company will have to strategically position itself very carefully, since China is quickly becoming a dominant player in exactly the verticals it wants to own (automotive, ML workflows, etc.). In other words, Nvidia needs the Chinese market for growth at the exact moment that door is slamming shut. How it navigates this challenge in the years ahead will determine much of its growth profile in the years ahead.

Rapid fire analysis

Short summaries and analysis of important news stories

Saudi Arabia’s Crown Prince Mohammed bin Salman. FETHI BELAID/AFP/Getty Images

US intelligence community says quantum computing and AI pose an ’emerging threat’ to national security – Our very own Zack Whittaker talks about future challenges to U.S. national security. These technologies are “dual-use,” which means that they can be used for good purposes (autonomous driving, faster processing) and also for nefarious purposes (breaking encryption, autonomous warfare). Expect huge debates and challenges in the next decade about how to keep these technologies on the safe side.

Saudi Arabia Pumps Up Stock Market After Bad News, Including Khashoggi Murder – A WSJ trio of reporters investigates the Saudi government’s aggressive attempts to shore up the value of its stock exchange. Exchange manipulation is hardly novel, either in traditional markets or in blockchain markets. China has been aggressively doing this in its stock exchanges for years. But it is a reminder that in emerging and new exchanges, much of the price signaling is artificial.

A law firm in the trenches against media unions – Andrew McCormick writes in the Columbia Journalism Review how law firm Jones Day has taken a leading role in fighting against the unionization of newsrooms. The challenge of course is that the media business remains mired in cutbacks and weak earnings, and so trying to better divide a rapidly shrinking pie doesn’t make a lot of sense to me. The future — in my view — is entrepreneurial journalists backed up by platforms like Substack where they set their own voice, tone, publishing calendar, and benefits. Having a close relationship with readers is the only way forward for job security.

At least 15 central banks are serious about getting into digital currency – Mike Orcutt at MIT Technology Review notes that there are a bunch of central banks, including China and Canada. What’s interesting is that the trends backing this up including financial inclusion and “diminishing cash usage.” Even though blockchain is in a nuclear winter following the collapse of crypto prices this year, it is exactly these sorts of projects that could be the way forward for the industry.

What’s next

More semiconductors probably. And Arman and I are side glancing at Yelp these days. Any thoughts? Email me at danny@techcrunch.com.

This newsletter is written with the assistance of Arman Tabatabai from New York