Deep Access

We’ve got some exciting news to share with you before we get into today’s AMA…

Next Wednesday, November 20, I’m going to be launching my first brand-new research product in three years.

It’s been a long road to this launch with plenty of speed bumps, courtesy of previous management. But this is a product five years – and more than $1 million spent in research and development – in the making.

And at last, it’s ready to roll.

It’s a new trading service we’re calling Deep Access… named for the proprietary, real-time dataset only my team and I have access to, with which we’ve developed our own deep-learning technology that allows us to predict future swings in equities.

Without giving too much away, this niche dataset offers us some insight into all short positions and trades, even those occurring in dark pools, on all equities.

This was a hard-earned dataset not available to investors. In fact, from what we’ve been told, we’re the only institution actively using it. It wasn’t easy to gain access to it, but now that we do, it has proven to be a gold mine for high-probability trades. This is the kind of access typically reserved for well-funded hedge funds and investment banks.

And we’re going to use this dataset to profit off of everyday volatility in equities…

The Deep Access AI is consistently analyzing millions of data points and predicting the fall in individual stock prices in both bull and bear market conditions. It is going to be an incredible tool for us to both profit from declining equities, as well as avoid stocks that are highly likely to fall.

If this sounds familiar, you might remember a similar technology informing our Neural Net Profits crypto advisory. Deep Access isn’t crypto-focused in the way Neural Net Profits is, and we’ll be trading in stocks likely to decline in even shorter time frames – within 30-40 days versus Neural Net’s 60-day windows.

If a fast-paced trading service sounds intimidating to you, don’t worry. During my event on November 20, there will be a demonstration of how you can execute these trades. It’s much more straightforward than you might expect.

I can’t wait to finally be able to share this service with you. If you’re at all interested or want to learn more, you can go right here to automatically sign up to join us next Wednesday, November 20, at 8 p.m. ET.

I hope to see you there. Now, on to the AMA…

Starlink Mobile Service Market Disruption

Hi Jeff,

I have a question about Starlink market disruption. My assumptions may be flawed, but the way I understand it, the next-generation Starlink satellites have much more capacity with direct-to-cell.

Being that it requires Starship to get to orbit, assuming Starship progresses timely, they are launching them in the next year or two, and have a network of these satellites in operation, how does that affect the mobile service provider industry?

Does SpaceX become a full-service provider or an MVNE? Does this technology mostly obsolete towers or is there just not enough capacity? It seems that SpaceX is on track to be a huge disruptor in that space.

 – Aaron W.

Hello Aaron,

I’m glad you asked about this as I’m sure many others have the same question.

SpaceX, through Starlink, is a broadband network provider as it has direct relationships with its end subscribers. Direct-to-cell, however, is different, at least for now. Starlink, as applied to direct-to-cell services, acts as an MVNE (mobile virtual network enabler).

The distinction is that Starlink provides the technical infrastructure to enable those direct-to-cell services but doesn’t own a direct relationship with mobile subscribers.

The Starlink strategy is to partner with mobile network operators (MNOs) around the world in each country to provide those direct-to-cell services.

Theoretically, SpaceX can become its own mobile network operator, but it is highly unlikely. Even with the expanded capacity of the direct-to-cell Starlink satellites, bandwidth will still be dramatically different compared to a 4G or 5G network performance.

And direct-to-cell will only work well when there is a direct line of sight to a Starlink satellite. It won’t work properly if indoors or in a major city with tall skyscrapers obstructing the line of sight, or even standing under an overpass. It is designed to provide network coverage where there is no terrestrial wireless network coverage.

So, to answer your question specifically, no, Starlink will not make cell towers obsolete.

It helps wireless network operators provide ubiquitous coverage in their geographic area. This is a powerful idea. Most countries have regulatory requirements to provide connectivity requirements for the entire population in exchange for using radio frequency spectrum.

The problem is that the more rural it gets, the less economical it is to build out a new wireless network. Partnering with Starlink will likely meet regulatory requirements and give 100% network coverage in any geographical area.

As for timing, the direct-to-cell Starlink satellites are already in orbit. SpaceX has been launching them on Falcon 9 rockets for some time now. In fact, the most recent Starlink direct-to-cell launch was on November 13, which you can see below.

If you’d like to see the whole launch, you’ll find it right here. The above launch on Wednesday included 20 Starlink satellites, 13 of which were direct-to-cell.

In the coming weeks and months, we’ll see mobile network operators around the world turning on their direct-to-cell services.

Sustainability in the Face of Compute Capacity Struggles

Hi Jeff,

Recently, you wrote, “It’s Not an AI Bubble, It’s a Race.” In light of current compute capacity challenges, what do you think of investing in AI infrastructure/GPU data center companies like Coreweave, Lambda Labs, Fluidstack, Ori, etc?

Allow me to provide a bit of color on my question please: Firstly, obviously, the question is not just about current compute capacity challenges, but whether and for how long that extends into the future. Also, whether GPUs will (continue to) be the go-to choice in the future or whether TPUs, LPUs, or other chips made by the multitude of up-and-coming chip designers (or even simply newer versions of A100/H200/GB chips) will become the new defacto architecture to run on. And if so, the costs of hardware turnover/upgrade relative to revenues.

On top of this, other competition will come as Google, MS, OpenAI, Meta, etc. further develop their own chips as well as sell off their own surplus (if any) compute capacity. Especially so, if we can (within the next 10 years?) get to an inflection point where the training of models reaches a critical mass, and no longer consumes the majority of compute resources.) I guess the question is how sustainable in the mid-long term (5-7 years), is the growth of these “GPU-as-a-service” companies?

One of the concerns I have is that I’m not really sure they have a defensible niche. Essentially (it seems to me), they’re just buying and selling, acting for the most part, basically as middlemen. And so they will be subject to all the market elements of arbitrageurs — demand/supply, price fluctuations, inventory and margins, hype/disillusionment cycles, etc.

I ask this question also, not just as/about an investment into companies in that sector itself, but also as compared to investing into say LLM operators (e.g., OpenAI, Anthropic, xAI) or into chip companies (like Cerebras, Groq, Lightmatter, Extropic, Rain). Or perhaps you believe in diversifying and investing in all three (or more) sectors of AI? Or do you personally have preferred options? Would be interesting to get your take on this…

 – Kim San L.

Hi Kim,

I appreciate all the context you provided around your questions. This is a very interesting topic and one that I could probably write a book about, so this will be a very compressed answer.

I also need to provide a reminder that my comments are meant to be general and aren’t meant in any way to be individualized investment advice. These are just some of my thoughts on the topics you raised.

For the benefit of all readers, first and foremost, when we are determining if an investment opportunity is a good opportunity, it is important to understand the enterprise valuation of the company that we’re analyzing.

It is undervalued? Is it overvalued? Is the valuation reasonable? And what is our outlook for the company as investors (i.e. will the valuation increase over a reasonable period)?

OpenAI just raised $6.6 billion last month at a $157 billion post-money valuation. Projected 2024 revenues for OpenAI are $3.7 billion, implying a valuation multiple of 42.4 times 2024 sales.

That’s a very rich valuation. For a slow-growing company, it could take years before the sales catch up with the valuation. The next raise could even happen at a down round (i.e. lower valuation).

But what if OpenAI’s revenues jump to $20 billion next year? Its forward multiple would be just 7.85 times 2025 sales. That’s attractive for a super-high-growth business, assuming that the company can operate at healthy gross margins and generate free cash flow.

This is just an example, but I hope it helps frame the rest of my thoughts.

Kim, I agree with you that private cloud service providers like the ones you mentioned are generally less attractive from an investment perspective compared to foundational AI software companies or AI-specific semiconductor companies.

Generally speaking, the latter two categories would have stronger competitive differentiation, better control over their product strategy, and ultimately better pricing power than cloud service providers that have to buy GPUs or other AI-specific semiconductors and compete with the likes of Google Cloud, AWS, Microsoft, Oracle, etc.

Scale is critical in those businesses, and yet, there will always be niche opportunities in cloud services that tend to be industry-specific. But in a market growing as much as AI cloud services, these cloud service providers can still be great investments.

Fluidstack and Ori are very small at the moment with limited funding, but Coreweave has raised more than $1 billion this year and is valued at more than $20 billion. Lambda Labs has raised $800 billion just a couple of months ago.

What typically happens in situations like this is the smaller players either carve out a good niche, grow, and ultimately get acquired by a larger cloud service provider, or they run out of capital and end up in an asset sale. Coreweave is on a path toward an IPO, Lambda Labs has a way to go.

NVIDIA and AMD GPUs have been the workhorses for AI training and inference because they are well-established, high-quality semiconductors that are widely available today. These products have been widely used in the industry for the last decade. They were the basic infrastructure for machine learning applications well before generative AI became all the rage.

It takes years for novel, AI-specific semiconductor designs to scale up to production levels needed for large-scale training and inference. And we should remember that all of these private companies are competing for manufacturing capacity at TSMC in Taiwan.

At this moment in the industry, companies building foundation models for AI aren’t very price-sensitive. They just need computational power as quickly as possible. And this is why GPUs are in such high demand.

We can be sure that demand will continue in the race for artificial general intelligence (AGI). The prize is so big, the first to get there will have an incredible competitive advantage. And investment will continue because several companies – not just one – will have AGI.

Once there are several AGI models, there will be a shift towards inference – the running of AI applications. This is where AI-specific semiconductors will benefit over the more traditional GPUs. The companies that you mentioned in this space – Cerebras, Lightmatter, Rain, Extropic, and Groq – are all excellent in this regard, and companies I follow closely.

Unfortunately, just about all the most exciting AI-specific semiconductor development is happening at private companies, which makes them hard to access for self-directed investors.

And to your point, Google, Meta, Microsoft, and Amazon have developed their own AI-specific chips to run their own applications and/or offer cloud services at improved gross margins with their own designs.

Self-directed investors can obviously invest in those companies, but the leverage to the semiconductors is tiny. What you’re really investing in is a cloud services business, an advertising business, a software business, or some combination of two.

Demand for both AI cloud services and AI-specific semiconductors will remain extremely high over the next five years. But we will see institutional capital leaning more heavily into those they believe will achieve greater scale.

We’ve already seen examples of this both in foundation models (i.e. the winners have already been chosen) and in cloud services.

To wrap things up, I strongly prefer companies with a strong point of product differentiation that can reduce friction in whatever problem they are trying to solve.

This accelerates adoption, growth, and ultimately valuation.

Can We Trust Republic?

Hi Jeff,

I am writing you today with concern on one of the VC Files presentation/Q&A with Andrew Durgee, the President of Republic, on 10/24. As we were a few weeks from the presidential election, a question came up about the future of Crypto depending on who won the election. His answer to this question, and I am paraphrasing here, was that there would be no material difference in the future of Crypto regardless of who wins. I found that answer to be absurd and apparently the Crypto market agrees as Bitcoin has hit $88,000 today.

My long-winded question is: How are we supposed to trust this company moving forward if the President answers a political question with such intellectual dishonesty?

 – David L.

Hi David,

I haven’t spoken with Andrew about this topic, so I can’t really comment on his position. But before his current position, he was the Head of Republic Crypto for many years.

There are two ways to look at the answer.

If we were to take a global view, even if Harris were to have won the election, the crypto markets would have still thrived… just not in the United States. But if we were to take the perspective of crypto just in the U.S., you’re absolutely right, there is a very material difference in outcomes between the two candidates.

Regarding the blockchain industry and digital assets, the worst four years have been under the Biden/Harris administration. The industry would be devastated had Harris won. For most people in the blockchain industry, this was a single-issue election. Trump’s reelection is extremely bullish for the domestic blockchain and digital assets industry.

From a global perspective, Andrew’s take wasn’t wrong. And I don’t want to conflate this perspective and the issue of whether or not we can trust Republic.

Republic is one of the most well-known and well-established companies in the crowdfunding industry. And the Republic Note is a digital asset backed by equity and assets of 623 companies right now.

In time, the value of the NOTE will be determined by the value of those underlying assets and the stream of dividends paid as the underlying companies are either acquired or go public.

Republic has been an innovator in the industry and the creation of the NOTE – a digital asset backed by equity from various companies – is an incredible innovation in the industry.

It’s like having one instrument to gain access to a large venture capital portfolio available to all self-directed investors.

An Ethereum Disconnect?

Hi Jeff,

It’s great to have you back! I’m an Unlimited member and I have a question regarding the 2 major cryptocurrencies.

In the past few weeks, we’ve seen Bitcoin rising and flirting with new highs. In the past, we would see Ethereum follow suit and hover around new highs as well. But not this time around. They are completely disconnected.

Do you have any insights as to why this is? Both of them have ETFs, and one would expect that with money pouring into them, it would affect Ethereum similarly to Bitcoin, but no. I really hope you can help me understand this issue.

Thanks, and keep up your great work.

– Frederick M.

Hi Frederick,

You raise an interesting point, especially after the results of the U.S. election, which is extremely bullish for cryptocurrencies and the blockchain industry.

To your point, why isn’t Ethereum running as well? It’s great news after all.

Historically, bitcoin has been the bellwether asset for the entire cryptocurrency industry. When it is running, it tends to be positive for the broader cryptocurrency market.

But from an institutional capital perspective, two things are often overlooked.

The first is that bitcoin makes up 60% of the total cryptocurrency market. This compares to 12.4% for Ethereum.

Bitcoin’s dominance is directly related to the dynamic that you’ve described. When we have a major breakout like the one that we have right now, it sucks liquidity out of the broader market. This has happened during every major cycle.

Understanding bitcoin’s dominance in relation to the broad digital assets market is a useful metric. It helps us understand how the market ebbs and flows throughout major multi-year cycles.

When we look at this metric over the last seven years, we can see what is getting ready to play out.

Bitcoin dominance is continuing to rise. It’s inching higher and higher. But as soon as it peaks out, ETH tends to be right next to its all-time highs.

The peak in dominance is met with ETH breaking through its all-time high. This causes ETH to seem like it’s lagging… until it goes into price-discovery mode.

We are nearing that moment right now.

If we look at the chart below, we can get a nice visual of what I’m referring to. The green boxes help us see when the price broke through the previous all-time high. The break is seen with the vertical green line.

The black line is Bitcoin Dominance. And we can see how quickly it drops as soon as ETH pushes past its all-time high.

And right now, we can see the price is inching back up towards its prior all-time high from 2021. Once this breaks, the rest of the market will light up.

But will ETH break new all-time highs? That’s a current question on the minds of some investors…

That’s because Ethereum hasn’t had a massive run like other digital assets. This negative sentiment has caused many to question how Ethereum handles gas fees, its transition to L2s, and a host of other niche topics – myself included.

The truth is sentiment like this is reflexive. Meaning when price is low or down, sentiment is bad. When price rises, everything is great. Price is a magical tonic that heals all wounds.

And if we dig into the recent flows taking place… along with certain signals… we should be confident that ETH will, in fact, make a run past its prior all-time high.

Ethereum ETFs have witnessed capital inflows since President Trump got re-elected. Before November 4, there had been a net outflow of capital out of Ethereum ETFs. It’s been a pretty terrible sign for institutional adoption.

But since the election, the net inflows since the ETF began to be traded are now positive. That’s thanks to the nearly $800 million of inflows happening over the last eight trading days. It’s historic.

What’s more, a major player acquired an infrastructure company that provides staking services for ETH. The company is Bitwise Asset Management, and it hints that demand for ETH’s yield is rising.

We’ve even witnessed other acquisitions such as Utopia Labs by Coinbase on November 13. Utopia Labs has built Ethereum-based tech that helps with onchain payment infrastructure that meshes with everyday bank-based checking accounts.

The flows, the acquisitions… This is all very recent. Interest is rising and looks like it won’t slow down. The market knows that the next four years at a minimum will be a drastically improved regulatory and economic environment for digital assets.

So while ETH’s price has been slow to make a move, it looks like history will repeat itself with another breakout move and new all-time highs.

This has been a great set of questions. Thanks to everyone who wrote in, and, as always, you can reach me or my team right here if you have your own comment or question. We read all the feedback you send in, just remember, I can’t give individualized investment advice.

And if you’re interested in getting Deep Access to ride out the coming wave of volatility in AI stocks, be sure to go here to automatically get your name on the attendees list for my event next Wednesday, November 20, at 8 p.m. Eastern.

Have a great weekend.

Jeff