Amazon Is Designing New Chips for Data Centers

Jeff Brown
|
Apr 12, 2021
|
Bleeding Edge
|
10 min read
  • The robotics trend is taking off…
  • Developers are doing great work with GPT-3…
  • Amazon is gearing up to make its own chips…

Dear Reader,

Every once in a while, I’ll read the annual reports or letters to shareholders of large incumbent companies.

It’s not because I’m interested in their business. While their businesses are often good, and they often pay a dividend, they simply don’t present the kinds of growth opportunities and investment returns that I can find in disruptive companies in the tech and biotech space.

This weekend, I took some time to read through the Letter to Shareholders from Jamie Dimon, Chairman and CEO of JPMorgan Chase.

I thought it would be fitting considering that this is a historic week in the financial services industry.

After all, financial technology (fintech) giant Coinbase is scheduled to go public on Wednesday. It is the “blue chip” representative of the cryptocurrency and blockchain industry.

Its last known valuation in the secondary markets was just above $100 billion. That’s a fraction of JPMorgan Chase’s $474 billion market cap – but just a hair away from Goldman Sachs’ $113 billion market cap. The delta won’t last long…

In light of the Coinbase IPO – and what is happening right now in the fintech space – one paragraph in the letter caught my eye:

Banks already compete against a large and powerful shadow banking system. And they are facing extensive competition from Silicon Valley, both in the form of fintech and Big Tech companies (Amazon, Apple, Facebook, Google and now Walmart), that is here to stay. As the importance of cloud, AI and digital platforms grows, this competition will become even more formidable. As a result, banks are playing an increasingly smaller role in the financial system.

I got a chuckle when I read this. Language like this is designed to paint a picture for regulators about how tough it is out there. The competition is all over the place… it’s not fair… they don’t play by the same rules…

In other words, JPMorgan Chase wants the regulators to focus on the competition, not on its own business. Meanwhile, JPMorgan Chase did just fine last year. It brought in record annual revenues of $122.9 billion and a net income of $29.1 billion.

No one is going to go hungry at JPMorgan Chase, but Dimon is feeling the heat. He knows what’s coming.

Coinbase, which is entirely focused on financial services around digital assets, will trade up on the IPO. I believe that we’ll see the company quickly trading in the $150–$200 billion enterprise value range.

And Stripe, another one of my favorite fintechs, was recently valued at $95 billion and growing. I doubt it will be too far behind Coinbase with its IPO.

While Walmart isn’t much competition for JPMorgan Chase, technology companies that are active in payments and fintech are. And their strength is growing.

They are literally rewriting the “code” for the financial system. And while their financial services may not yet be as big as JPMorgan Chase, their margins are much better, they move faster, and they aren’t very far behind.

I wish my friends at JPMorgan well. And I’ll be keeping a close eye out for the peak of its business. It won’t be long before the competition makes it too hard to sustain even a modicum of growth.

And to be fair, I look for the same inflection points at industry incumbents like IBM, Microsoft, and Intel.

We’re at an exciting inflection point in the financial services industry. The things that have been historically valued by customers and institutions are losing their importance. Consumers and many institutions simply want technology and services that are simple to use, faster, cheaper, and better.

And for many, not having a “trusted” intermediary in the middle charging high fees is exactly what the market wants.

Coinbase, Stripe, Plaid, Square, Robinhood, and the part of the blockchain industry focused on decentralized finance (DeFi) are the future. And the future is coming much faster than we think…

Now let’s turn to today’s insights…

The 24-hour worker that doesn’t get tired…

Boston Dynamics recently announced the newest addition to its suite of mobile robotics – Stretch. It’s a robot that’s designed to lift and move boxes within a warehouse facility.

Boston Dynamics is a company we’ve talked about several times in The Bleeding Edge. Readers may remember the company’s first commercialized robot named Spot. It’s a dog-like robot that can navigate its environment, carry objects around, and even open doors.

Boston Dynamics’ Spot

Source: Boston Dynamics

The technology is incredible. It is actually quite a breakthrough that incorporates bleeding-edge robotics with computer vision and forms of artificial intelligence (AI). But admittedly, the market applications have been somewhat limited.

As a reminder, this company was originally spun out of MIT in 1992. Then it was acquired by Google X in 2013. Google X sold it to SoftBank in 2017. And then Hyundai Motor Group acquired an 80% share of the company last year for $880 million.

It’s safe to say the 29-year-old company has made the rounds. It is difficult to execute a consistent strategy with all these changes.

But the launch of Stretch signals that Boston Dynamics is now focused. This is the most commercially viable robot that has significant scale to date.

Equipped with machine vision software, Stretch is designed to be integrated into any warehouse or logistics facility. It can move 800 boxes an hour using its mobile and dexterous arm.

And get this – the suction cups at the end of its arm allow Stretch to lift up to 50 lbs. with speed and mobility.

Have a look:

Boston Dynamics’ Stretch

Source: Boston Dynamics

As for performance and speed, Stretch is already on par with human workers… except the robot can work 24 hours a day. No need for food, drink, or rest. And Stretch can’t get injured.

That makes Stretch a no-brainer for many of the 150,000 warehouses and distribution facilities operating around the world. I expect we’ll see rapid adoption in the industry.

And it doesn’t stop there.

A smaller version of Stretch could be used to make package deliveries more efficient. I envision an autonomous delivery van pulling into a neighborhood with a smaller Stretch robot in the cargo area.

When the van parks, Stretch can work in unison with mobility robots like a larger version of Spot or the Digit robot from Agility Robotics to sort packages and deliver them door-to-door. No human oversight required.

What we are seeing here is the beginning of a massive trend. As we can see with Stretch, robotics technology is advancing at a rapid pace. And robots will soon become commonplace in our everyday lives.

And bleeding-edge robotics companies like Boston Dynamics will present us with massive investment opportunities in the coming years as the robotics space takes off.

GPT-3 is on the rise…

OpenAI just gave us a glimpse into how widespread GPT-3’s deployment has been… and it’s incredible.

We have talked about OpenAI and GPT-3 several times before in these pages. To bring new readers up to speed, OpenAI started as a nonprofit research institute in 2015. Then it spun out a commercial arm in 2019.

And GPT-3 is OpenAI’s prized product. This is an AI-enabled text generator that has Silicon Valley humming.

At the simplest level, GPT-3 can complete our sentences as we type. But that’s only the start…

GPT-3 has written entire technical manuals. It can write short stories on command with only a small prompt like, “Bob went to the carnival and won a stuffed animal for his daughter.”

And most recently, Microsoft Word integrated GPT-3 into its software. As users type in Word, the AI will offer to complete the sentences. And the more it gets to know each individual user, the better the predictions will be.

And OpenAI’s latest announcement reveals that GPT-3 has expanded far beyond Microsoft Word. In fact, tens of thousands of developers have built more than 300 applications using the technology.

Across all these applications, GPT-3 is producing 4.5 billion words every day. And it learns with every new word generated.

As such, GPT-3 will get better and better as developers use it and improve the code over time. And that means we are going to see more and more practical uses for GPT-3.

One area that is ripe for disruption is customer service. GPT-3 is already being used to analyze customer feedback and sentiment. It can also create help desk tickets for customers who need support.

And it can provide insight to help companies better understand their customers’ needs and wants.

Companies get more efficient customer service without needing to expand their customer service departments. It makes perfect sense.

A company called Algolia provides another great example of GPT-3’s value.

Algolia uses GPT-3 in its new web search product, Answers. The service connects customers to specific content that can answer their questions. Customers simply type in their question and get directed to the proper answer.

After integrating GPT-3 into this process, Algolia saw a 91% increase in the service’s ability to answer questions accurately. That shows us how smart GPT-3 has become.

This is a tremendous leap in terms of usefulness. And it demonstrates the power of GPT-3.

Looking forward, I can envision GPT-3 even competing with Google’s search engine. Google’s reign won’t last forever.

And perhaps the biggest thing we’ve learned is that Microsoft’s exclusive license doesn’t prevent other developers from working with GPT-3. That’s fantastic news. I was very worried about this back in September when that licensing deal was struck.

It turns out that Microsoft simply has the exclusive right to look at the source code and modify it if it chooses. That’s not nearly as bad as I feared.

So we will continue to watch GPT-3 closely from here. Adoption is ramping up, and I expect this product will generate significant revenue for OpenAI. And it will also lead to some exciting new consumer applications that will prove to be extremely useful.

Amazon is jumping on this massive trend…

We’ll wrap up today with some far-reaching developments at Amazon. The tech giant is gearing up to improve its internal infrastructure and Amazon Web Services (AWS) division.

To accomplish this, Amazon is now designing custom semiconductors for network switching. We can think of these as the brains of ethernet switches. They help get data from point A to point B in Amazon’s massive data centers. The switches help the network respond to varying demands throughout the day. They help the network efficiently use its resources.

Here’s what’s interesting about this… This is a space that Amazon has traditionally relied on merchant silicon for. These are semiconductors produced by other companies. Amazon has largely relied on Broadcom for this technology.

This is where Amazon’s 2015 acquisition of Annapurna Labs comes in. Amazon shelled out $350 million for this early stage semiconductor company, and now we know why. Amazon has been planning to design chips for network switching in-house for years now.

And this in-house design will allow Amazon to customize the chips to mesh perfectly with the software used in its AWS business unit. Clearly, Amazon feels that it can deliver better performance with its own designs.

That said, Amazon won’t manufacture its own chips. It will rely on Taiwan Semiconductor Manufacturing Company (TSM) for that. This is a company that Near Future Report subscribers know well. We are sitting on gains of over 225% on TSM to date.

And this is all part of a massive trend unfolding in the industry.

Big tech companies like Apple, Google, Microsoft, and now Amazon are each committed to designing their own semiconductors. It’s all about customization and optimization in order to improve overall performance.

It’s also a message to the semiconductor industry that their products aren’t delivering the level of performance that they need. Otherwise, they wouldn’t look to design and develop in-house.

Historically, these hyperscale companies have focused their semiconductor design efforts on very specific AI and machine learning applications.

But Amazon is now targeting a more general application like network switching. While this is bad news to certain semiconductor companies, it is good news to certain semiconductor companies that license intellectual property for semiconductor design.

This is a major trend to be aware of in the semiconductor industry. And we can be sure that this move by Amazon will only make AWS an even more dominant player in the cloud services industry.

Regards,

Jeff Brown
Editor, The Bleeding Edge

P.S. OpenAI’s technology is disrupting the customer service field… taking customer service from analog processes to digital ones.

And now I’m seeing a similar progression start to unfold in an industry set to experience tremendous growth as it goes digital.

It’s an event I call the digital leap.

When it unfolds, we see once-in-a-generation investment opportunities. I’m talking about companies like Amazon and Netflix. Early investors in these companies made an absolute fortune by investing when these industries were making a digital leap.

That’s why I’m so excited to see the massive $11.9 trillion health care industry preparing for this same kind of digital leap. Up until now, it’s been held back from innovating by red tape and bureaucracy.

But thanks to the COVID-19 pandemic, we’re finally seeing the transition begin. And that means the time to take a position in this industry is now.

If you want to learn more about what I believe could be the last digital leap… and what investors need to do to prepare, click here.


Want more stories like this one?

The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.