May 26, 2019
Category archive


Hailo launches its newest deep learning chip

in AI chips/Artificial Intelligence/ceo/Companies/Delhi/hailo/India/integrated circuits/Intel/Mobileye/neural networks/nvidia/Politics/Technology/Tel Aviv by

Hailo, a Tel Aviv-based AI chipmaker, today announced that it is now sampling its Hailo -8 chips, the first of its deep learning processors. The new chip promises up to 26 tera operations per second (TOPS) and the company is now testing it with a number of select customers, mostly in the automotive industry.

Hailo first appeared on the radar last year, when it raised a $12.5 million Series A round. At the time, the company was still waiting for the first samples of its chips. Now, the company says that the Hailo-8 will outperform all other edge processors and do so at a smaller size and with fewer memory requirements. “By designing an architecture that relies on the core properties of neural networks, edge devices can now run deep learning applications at full scale more efficiently, effectively, and sustainably than traditional solutions, while significantly lowering costs,” the company explains.

The company also argues that its chip outperforms Nvidia’s comparable Javier Xavier AGX in some benchmarks, all while using less power and hence running cooler — something that’s especially important in small IoT devices.

We’ll have to see if that works out in practice once more engineers get their hands on these chips, of course, but there can be no doubt that the demand for AI chips on the edge continues to increase. A few years ago, after all, the market shifted away from a focus on centralizing all processing in the cloud to moving to the edge, in an effort to improve latency, reduce bandwidth cost and provide a more stable platform that doesn’t depend on network connectivity.

Like Mobileye before it (which was later acquired by Intel), Hailo is working with OEMs and tier-1 suppliers in the automotive industry to bring its chip to market, but it’s also looking at other verticals, including smart home products and really any industry where a high-performance AI chip is needed for object detection and segmentation, for example.

“In recent years, we’ve witnessed an ever-growing list of applications unlocked by deep learning, which were made possible thanks to server-class GPUs,” said Orr Danon, CEO of Hailo. “However, as industries are increasingly powered and even upended by AI, there is a crucial need for an analogous architecture that replaces processors of the past, enabling deep learning to run devices at the edge. Hailo’s chip was designed from the ground up to do just that.”

Internet connectivity projects unite as Alphabet spinout Loon grabs $125M from SoftBank’s HAPSMobile

in 5g/aerovironment/airbus/Alphabet/aquila/Australia/Companies/Delhi/Facebook/Google/India/internet access/internet service providers/Loon/nvidia/peru/Politics/puerto rico/social network/SoftBank/SoftBank Group/TC/telecommunications/United States/Vision Fund/Vodafone/wirecard/X by

Two futuristic projects are coming together to help increase global internet access after Loon, the Google spinout that uses a collection of floating balloons to bring connectivity to remote areas, announced it has raised money from a SoftBank initiative.

HAPSMobile, a SoftBank project that is also focused on increasing global connectivity, is investing $125 million into Loon, according to an announcement from SoftBank made this morning. The agreement includes an option for Loon to make a reciprocal $125 million investment in HAPSMobile and it includes co-operation plans, details of which are below.

HAPSMobile is a one-year-old joint venture between SoftBank and U.S. company AeroVironment . The company has developed a solar-powered drone that’s designed to deliver 5G connectivity in the same way Facebook has tried in the past. The social network canceled its Aquila drone last year, although it is reported to have teamed up with Airbus for new trials in Australia.

Where Facebook has stumbled, HAPSMobile has made promising progress. The company said that its HAWK 30 drone — pictured below in an impression — has completed its initial development and the first trials are reportedly set to begin this year.

Loon, meanwhile, was one of the first projects to go after the idea of air-based connectivity with a launch in 2013. The business was spun out of X, the ‘moonshot’ division of Alphabet, last year and, though it is still a work in progress, it has certainly developed from an initial crazy idea conceived within Google.

Loon played a role in connecting those affected by flooding in Peru in 2017 and it assisted those devastated by Hurricane Maria in Puerto Rico last year. Loon claims its balloons have flown more than 30 million kms and provided internet access for “hundreds of thousands” of people across the world.

In addition to the capital investment, the two companies have announced a set of initiatives that will help them leverage their collective work and technology.

For starters, they say they will make their crafts/balloons open to use for the other — so HAPSMobile can tap Loon balloons for connectivity and vice-versa — while, connected to that, they will jointly develop a communication payload across both services. They also plan to develop a common ground station that could work with each side’s tech and develop shared connectivity that their airborne hardware can tap.

Loon has already developed fleet management technology because of the nature of its service, which is delivered by a collection of balloons, and that will be optimized for HAPSMobile.

The premise of HAPSMobile is very much like Loon

Outside of tech, the duo said they will create an alliance “to promote the use of high altitude communications solution with regulators and officials worldwide.”

The investment is another signal that shows SoftBank’s appetite in tech investing is not limited to up-and-coming startups via its Vision Fund, more established ventures are indeed also in play. Just yesterday, the Vision Fund announced plans to invest $1 billion in German payment firm Wirecard and its past investments include ARM and Nvidia, although SoftBank has sold its stake in the latter.

Claire Delaunay will be speaking at TC Sessions: Robotics + AI next week at UC Berkeley

in Anthony Levandowski/Artificial Intelligence/california/Claire Delaunay/colin angle/Delhi/Events/Google/India/marc raibert/Melonee Wise/nvidia/otto/Politics/robotics/TC/TC Sessions: Robotics + AI/TC Sessions: Robotics+AI 2019/transport/Uber by

We’re a week out from our third-annual TC Sessions: Robotics event, and we still have some surprises left to announce. I know, we’re just as surprised as you are. We’ve already announced that Marc Raibert, Colin Angle, Melonee Wise and Anthony Levandowski will be joining us in Berkeley next week, and today we’re adding Claire Delaunay to the list of distinguished names.

Delaunay is VP of engineering at NVIDIA. Prior to NVIDIA, she worked as the director of Engineering at Uber, after the ridesharing service acquired her startup, Otto. She has also worked as the robotics program lead at Google.

She is currently the head of NVIDIA Isaac. The company’s robotics platform is designed to make it easier for companies of various experience levels and means to develop robots. Delaunay will discuss the platform and showcase some of NVIDIA’s in-house robotics reference devices, including Kaya and Carter.

Speaking of NVIDIA, TechCrunch is partnering with them on April 17 (the day before the conference) to host a Deep Learning for Robotics workshop at UC Berkeley. This in-person workshop will teach you how to implement and deploy an end-to-end project through hands-on training in eight hours, led by an instructor. Click here to learn more about the workshop.

Hear from Delaunay and other awesome speakers next week at TC Sessions: Robotics + AI. Purchase your $349 tickets now before prices go up $100 at the door. Student tickets are just $45 — book yours now.

EC Weekly: Gaming, crypto, shipping and the multiple future strategies of tech

in accessibility/Alphabet/Amazon/Cooley LLP/Delhi/Facebook/Google/India/Larry Page/Microsoft/niantic/Nintendo/nvidia/Politics/Sony/TC by

Niantic EC-1

Illustration by Nigel Sussman

Greg Kumparak published the first part of his planned four part EC-1 series on Niantic yesterday, focusing on the founding story of the AR/gaming unicorn from Keyhole and Google Earth to a complicated spinout from Alphabet. Lots of great nuggets on how companies get formed and built, but one I particularly enjoyed was this one:

Like most companies, Google doesn’t like when employees leave. Especially employees who ran key parts of the company for years. Leaving means competition. Leaving means potential opportunities lost.

John [Hanke, CEO of Niantic] eventually sat down with Larry Page to figure out what it’d take to keep him within Google . They talked about John’s interest in augmented reality. They talked about a book called Freedom™ by David Suarez, which centers around an out-of-control AI that taps a network of real-world operatives to control the world (the earliest hints of Niantic’s first game, Ingress, already sneaking in here years before it’s made.)

John wanted to take his interest in AR and his background in maps and gaming and mash them all up and see what it could look like. Larry wanted it to happen within Google.

What I loved is that Eliot Peper wrote a piece for Extra Crunch just a few weeks ago about the importance of speculative fiction in the creation of startups, and also gave a guide on just what books he recommends to find your next startup.

Expect Part 2 of the Niantic EC-1 to drop early next week as we do a rolling release.

Game streaming is the new battlefield among tech giants

Bryce Durbin / TechCrunch

Game streaming is quickly becoming one of the most important strategic arenas for owning users, with offerings from all major tech and gaming companies. Devin Coldewey provided a comprehensive strategic overview of the stakes involved this week, and why so much money is being poured into a technology that until now seemed impossible due to bandwidth and latency. It’s like Super Smash Bros: Tech Melee edition:

NVIDIA and OpenAI’s capped returns

in Ant Financial/Asia/Delhi/Government/Hardware/India/LinkedIn/maimai/nvidia/OpenAI/Politics/The Extra Crunch Daily/Zhaopin by

Editor’s note: Starting as a trial, the Extra Crunch Daily newsletter is going to be delivered Tuesday-Saturday, in order to faithfully analyze the happenings in the startup and financial world Monday-Friday.

Open AI’s capped returns

OpenAI announced yesterday that they are going to be offering a “capped return” security for investors as part of the for-profit/non-profit split the organization is creating:

As mentioned above, economic returns for investors and employees are capped (with the cap negotiated in advance on a per-limited partner basis). Any excess returns go to OpenAI Nonprofit. Our goal is to ensure that most of the value (monetary or otherwise) we create if successful benefits everyone, so we think this is an important first step. Returns for our first round of investors are capped at 100x their investment (commensurate with the risks in front of us), and we expect this multiple to be lower for future rounds as we make further progress.

I candidly don’t understand this structure at all. For venture capitalists — and particularly early-stage investors — returns are driven by one, maybe two, and extremely rarely three startups in a portfolio (that would be Benchmark’s 2011 fund, which includes Uber, Snap, and WeWork). That one outlier investment may drive a majority of all fund returns. If OpenAI were to be that investment, how you could you possibly relinquish the remaining upside? Maybe you could prospectively sort of accept this, but how would you explain to LPs that “ah, yes, seven years ago we decided to give up that next 150x” or whatever.

OpenAI LP (the for-profit entity) is trying to target more mission-oriented investors, who presumably value incentive alignment but not (huge) profits. That’s fine, but the idea of capping a return as a mechanism to capture run-away value creation seems really off to me and should be discouraged.

My colleague Devin Coldewey also had a negative take, but sort of in the opposite direction — that OpenAI “may not be quite so open going forward” and is going to focus more on profits than science. That’s a fair criticism as well, although I think the profit motive will get us to AGI faster.

You’re reading the Extra Crunch Daily. Like this newsletter? Subscribe for free to follow all of our discussions and debates.

With Mellanox deal, NVIDIA buys a chance to salvage its growth

Photo by David Becker/Getty Images

Written by Arman Tabatabai

NVIDIA confirmed whispers Monday when it announced it was acquiring adjacent semiconductor player Mellanox for $6.9 billion. Mellanox specifically focuses on interconnects and networking components that transfer data between cloud compute and storage resources.

The strategic rationale for NVIDIA is fairly straight-forward despite being a little outside of the company’s core competency. As we’ve discussed a few times before, NVIDIA got absolutely crushed towards the end of last year as the company struggled to find growth while facing headwinds from a dried up crypto market, a testy geopolitical backdrop, customer erosion, and increased competition. NVIDIA cuts its sales guidance by $500 million in the last quarter which, as the NYT pointed out, CEO Jensen Huang called “a real punch in the gut.”

NVIDIA has been betting the farm on diving into the data center, cloud computing, and supercomputer/AI markets that require parallel computation well served by NVIDIA’s graphical processing unit (GPUs). With Mellanox, NVIDIA will not only gets access to a segment with higher margins than its current operations but will, more importantly, be able to offer solutions across the full compute stack for data storage and AI/ML.

As TechCrunch’s Ingrid Lunden put it:

“While NVIDIA has focused its energies on computing, Mellanox works across Ethernet and other networking technologies — complementary areas for the two when addressing new computing and data transfer challenges brought about with the rise of AI, cloud services, an explosion of smartphone and other connected device usage and as-yet nonexistent tech like self-driving cars, which will put even more strain on our data infrastructure.”

The deal had been fairly well-telegraphed prior to the official announcement and is expected to be cash and earnings accretive. And the purchase price doesn’t appear to be too outlandish either — especially given a bidding process Huang described as “very competitive” — coming in slightly below the over $7 billion NVIDIA was rumored to be offering in order to outbid Intel, Xilinx, and Microsoft, all of whom had been linked as potential buyers during the past year in which Mellanox has reportedly been up for sale .

Notably, Intel seems to have missed out again here during a time where the company has been pouring money into R&D trying to play catch-up after struggling in recent years to keep up with the industry’s transition to new technologies.

NVIDIA stock was up around 7% on the day and Mellanox traded up to roughly $118 — just below the $125 per share acquisition price — with the market seemingly baking in a five-to-six percent chance of the deal not going through given the US government’s increased scrutiny on the global chip industry and pushback seen in prior semiconductor transactions. While a rejection of the deal would certainly be negative for NVIDIA, the company would only have to cough up a termination fee of $225-$350 million if the deal is blocked by shareholders or regulators and both leadership teams seem to be on board.

For NVIDIA, it seems like a small price to pay for a new shot at growth and a chance to quickly gain share in an increasingly competitive market.

Where is China’s new NASDAQ?

Photo by JOHANNES EISELE/AFP via Getty Images

China has a money problem (well, it has a lot of money problems, but let’s just focus on one for today). The country has produced a dizzying array of global-scale technology companies, including Alibaba, Tencent, and many more. The problem is that these startups grow up in China, but perform their IPO debuts overseas, typically in New York and also often in Hong Kong. There are a whole lot of reasons why this happens, but it annoys the hell out of the senior Chinese leadership.

So the Shanghai Stock Exchange, one of the two leading markets in the country, has been working with regulators to introduce a “NASDAQ-style” trading board that would have fewer rules on new issues. Those more lenient rules would include allowing companies to be unprofitable at IPO and to allow for multiple share classes, presumably with differential voting rights. In other words, they are designed for Silicon Valley-style startups.

We learned last week that the board’s introduction will come near the end of May, and it unveiled a nearly final set of rules for the new exchange last week. That’s months late though, since back in December, the exchange had said that new equity issues could begin trading as early as March.

The reason all of this minutia matters is because of Ant Financial . The Chinese fintech company was last valued at $150 billion, and its IPO, which has been rumored for months now, will be one of the major financial blockbusters of the year.

Where Ant Financial chooses to debut is a hugely important question for these exchanges, and for getting a read on the future divide between U.S. and Chinese capital markets. At its scale, it could almost single-handedly christen Shanghai’s new board, and indeed, it is rumored that the company wants to do just that. Certainly the Chinese government wants the company to trade locally.

So the question is whether it has the time to wait for Shanghai to get all of its pieces in order, while also ignoring the large capital markets in New York, London, and Hong Kong that would almost certainly have to be tapped for a company its scale.

LinkedIn’s failures in China

Illustration by Bryce Durbin/TechCrunch

It’s not every day you get a direct takedown of a product by that product’s former leader. But over the weekend, former LinkedIn China president Derek Shen blasted the company’s approach to China, according to a translation by Jill Shen at TechCrunch editorial partner TechNode (who I presume is unrelated):

“It’s horrible that the LinkedIn product managers don’t even realize they have lagged way behind a list of new social networking services such as WeChat, feeling good about themselves instead,” said Shen in a LinkedIn post on Monday. The former LinkedIn executive said that he tried to improve the platform when he joined the company six years ago, but struggled to make progress as it involved so many stakeholders within the organization.

(Of course, knocking LinkedIn’s product is a favorite pastime of pretty much any worker in Silicon Valley today).

LinkedIn first took China seriously in early 2014, and has had reasonable success in the interim, growing to around 41 million users. LinkedIn is unique among Western-run social networks in having (any) access to the Chinese market — essentially no other major network (including Twitter and Facebook) has passed through the Great Firewall.

Yet, its fortunes appear to be turning. LinkedIn, which is owned by Microsoft, is feeling a bit of a pincer from both Chinese and Western critics. The professional network has followed the censorship edicts of Beijing, much to the chagrin of human rights organizers. It has also added in a real name requirement linked to mobile phone numbers, which is now mandated by the government.

Meanwhile, domestic competitors like Maimai (脉脉) and Zhaopin (招聘) are building traction with more native products, to Shen’s point above. Maimai in particular has raised hundreds of millions in venture capital and is rumored (like all late-stage companies) to be targeting an IPO.

We talk a lot about the market-entry barriers that China’s government has placed on Western tech companies, but at least when it comes to consumer apps, it is also important to note that product cultural awareness doesn’t come instantly. Even if China’s markets opened tomorrow, these apps would still have to compete in the marketplace, and there is no guarantee that Chinese professionals want garbage InMail offering “growth services” any more than Silicon Valley workers do.

Eliot Peper and “narrative responsive design” on the web

Novelist and strategist Eliot Peper gave Extra Crunch readers a lengthy reading list of great speculative fiction a few weeks ago to help inspire the creation of startups. Now, one of his major projects has been published.

A few years ago, Peper published True Blue, a short story about discrimination in which people’s life outcomes are determined by the color of their eyes. It’s a parable to our own world, infested with the kind of speculative details that Peper is known for.

After publishing the short story, he teamed up with Phoebe Morris and Peter Nowell to bring a fully-illustrated and responsively-designed version of the story to life, with some funding from TechStars founder David Cohen.

What’s quite exciting about this project is seeing how artists are using the web as a deeper narrative platform. From Peper’s discussions of how the team made the product:

One of the counterintuitive lessons we learned was how powerful it is to obscure certain details, letting readers bring more of their imagination to the story. Specifically, we discovered that detailed lines often trigger the sense of something being depicted for you, so we smudged and faded and shadowed until we felt the right balance of detail and suggestion. This philosophy carried through to design — which so often aims to reduce tension by making experiences simple, intuitive, and convenient. But stories thrive on conflict, and Peter challenged himself to use design to evoke tension instead of erasing it.

He even engineered a new tool that cropped images so that they adapted to different devices and screen sizes not only by changing size, but actually changing image composition to preserve narrative content and emotional impact. When I told him about the project over a slice of Arizmendi pizza, author/friend/media experimenter Robin Sloan coined a term for this new technique: Narrative Responsive Design.

A lot of work yes, but the wait and effort I think are worth it. Read the story and learn more about the process of making it.

Editor’s Note

  • We are slowing down a bit on the infrastructure side that we have been discussing ad nauseam.


To every member of Extra Crunch: thank you. You allow us to get off the ad-laden media churn conveyor belt and spend quality time on amazing ideas, people, and companies. If I can ever be of assistance, hit reply, or send an email to

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

You’re reading the Extra Crunch Daily. Like this newsletter? Subscribe for free to follow all of our discussions and debates.

1 2 3 7
Go to Top