It’s AMA day…
We’ve got a great batch of questions to get through today. Thanks to everyone who wrote in, and if you have your own question you’d like to see answered in one of our Q&A-style Bleeding Edge issues, you can write us right here.
On today’s docket, we’ve got questions about whether or not the data center buildout is the most pragmatic use of existing energy resources (hint: it is)… a perceived threat to American startups… and how Brownstone Research plans to compete with the rise of AI-powered investment tools.
We’ve also got a fun peek into the production process behind the featured images you see at the top of every Bleeding Edge.
Let’s dive in…
So Jeff….
I just read today’s letter about the Bismark conference and the massive amounts of power projected to be required in the near future for all these emerging “data centers”, and I have to ask…
Is this really the most efficient, prudent, and/or economic use of the consumption of our existing (fossil fuels) energy sources??? What’s the payoff if we lose our ability to heat homes, drive our ICE vehicles, make our plastics, etc.?
If five years ago we were projected to have a 100-year supply of fossil fuels… what will it be now if we direct so much of that energy to data centers, bitcoin mining, or other what might seem to some as non-essential to the continuation of human life?
Love to hear your thoughts on this.
– John N.
Hi John,
I’m glad you asked about this as there has been a radical shift in thinking over the past couple of years.
Before this period of acceleration that started in late 2022, there was so much virtue signaling about efforts towards carbon neutrality, net zero, electrification (because there are no emissions), and offsetting electricity consumption with carbon credits.
Things have completely changed.
Now it’s more about increasing energy consumption at all costs, as quickly as possible. To your point, energy demands are increasing far more quickly now that the industry realizes we need to build $100 billion-plus AI data centers and eventually $1 trillion data centers. Efforts have quickly shifted away from reducing energy consumption.
Semiconductor companies are making successful efforts every year to improve the energy efficiency of semiconductors with each generation of semiconductor technology. But the nuance is that the energy efficiency improvements are per unit of compute.
Just because the chips are more energy efficient, it doesn’t mean energy consumption is being reduced. Quite the opposite, energy consumption is rapidly increasing because more of these chips are being used to take advantage of the higher performance.
This all might sound like it’s going in the wrong direction, but what I like about it is that it has forced the industry to be more honest and pragmatic about energy production.
Here are a few things that are now openly acknowledged:
The discussion is much more honest now.
And as a result, we’re seeing a resurgence of support for nuclear power, both fission and fusion, as the only way to meet the energy demands required for this massive data center infrastructure buildout.
I wrote about the sentiment shift surrounding nuclear fission not long ago in The Bleeding Edge – Washington’s Radical Shift Toward Nuclear. Nuclear power is the only option capable of gigawatt-scale clean energy production 24/7/365… and we need that power boost to support these massive data center facilities.
Do we need all the new data centers? I would argue emphatically yes. And my position is based on my prediction that we’ll develop artificial general intelligence (AGI) before the end of 2026.
AGI is critically important because it will be the largest productivity leap in history. AGI will be capable of self-directed research in any discipline. This will result in breakthroughs in just about every field we can imagine.
For example, AGI will be critically important for operating nuclear fusion reactors, which I believe is the best path forward for a decentralized, clean, limitless, and cheap energy source capable of providing for the world’s energy needs.
AGI will help us solve for cheap, clean energy and rapidly reduce our reliance on fossil fuels. It will lead to a world of abundance, which is why this period of effective acceleration (E/ACC) is so important. We just need to expend a little more energy to get there.
Jeff, would you be willing to comment on this article published 9/4 on Marketwatch titled “These investors are killing the game-changing tech startups of America”?
– Bart M.
Hi Bart,
This is an interesting topic because it gives us a chance to explore the shifting priorities of the venture capital and private equity industries over the last few years.
For the benefit of our readers, the article makes one key point to argue its position:
Almost 90% of America’s 120 publicly traded unicorns – those valued at $1 billion or more before they went public – were unprofitable in 2023.
The article’s authors also make the point that entrepreneurship has become a lifestyle. Because of this, many entrepreneurs who aren’t likely to succeed are getting funded.
This is definitely true, especially in places like New York and Silicon Valley. Way too many weak ideas and founders get funded due to the glut in available capital looking to the private markets for better returns.
And the reality is that most startups fail, but out of those that succeed there are always a handful that deliver extraordinary returns that make up the majority of a venture capital fund’s returns. That’s the nature of venture investing.
But back to the above point about 90% of those unicorns being unprofitable in 2023. The authors argue that this is a bad thing and that VCs are “killing the game-changing tech startups of America.”
This is complete nonsense.
It suggests that they would feel differently if 90% of those companies were profitable as if profitability is the metric that matters most when going public. This is foolish.
For example, Tesla (TSLA) went public in 2010 at a valuation of $1.58 billion (unicorn status). It was extremely unprofitable at the time.
In fact, it remained unprofitable through 2019 – 16 years after its founding. Now, Tesla is worth $716 billion, has a net cash position of $18.2 billion, and will generate $2.1 billion in free cash flow this year.
I could provide example after example just like this over the last 20 years.
Companies go public to access new capital and provide liquidity for existing investors. They don’t need to wait until they are profitable to do so. And as long as a company (public or private) can raise additional investment or debt to get it to the stage where it is generating free cash flow, then the business will be sustainable as an independent company.
And if a company can’t make it to free cash flow but still has value, the company will likely be acquired by a larger company that can profitably better leverage those assets.
VCs haven’t failed if they help take a portfolio company public while it’s still unprofitable. Their priority is simple – to increase the value of their portfolio companies over time and provide a compelling rate of return for their limited partners through some kind of liquidity event (acquisition or IPO).
There has recently been a shift in mentality though.
Up until 2021, the VC community was leaning in on “growth at all costs”. What this means is that VCs were willing to fund unprofitable businesses that continued to lose money as long as the revenue growth remained high. The idea is to scale as fast as possible and then monetize farther down the road. Profits can come later.
When interest rates are low and capital is cheap, this actually works well. But when interest rates are high, capital is expensive, and investors become more risk averse… things change.
Since 2022, we’ve seen a much stronger focus on the path to free cash flow breakeven. There is still the expectation of growth, but private companies now need to demonstrate a business plan that has a clear path to profitability in a shorter timeframe.
The venture markets are much more pragmatic now since 2022 and I don’t expect that will change until interest rates drop significantly – probably 250–300 basis points – and we see a broad-based bull market in tech/biotech.
And just to finish on a very positive note, there are a ton of game-changing startups that have been founded in the last 10 years that are still private and just waiting for better market conditions and timing to access the public markets.
I’m very excited about what’s to come. Once interest rates come down, and the market conditions improve, there will be hundreds of exciting IPOs. The pent-up demand and backlog are incredible.
And I can assure you that the best ideas and the best founders have no trouble at all raising capital in the current environment. The “game-changing tech startups” are getting all the funding they need and are actively building for the future.
Hi Jeff,
With the exponential advancement of AI, have you started researching the future of investing when AI takes control of all markets? How will Brownstone Research outsmart an AI investing agent that can “instantly” trade 24/7/365 in any market anywhere?
Unfortunately, I can see a point in the very “near future” where financial services such as Brownstone Research could be replaced by an AI “investment research” agent. How will Brownstone Research beat “the market” when AI is running it? Will national governments ban “AI trading”? What is the future of trading/investing in a world where AI can immediately trade an arbitrage?
I’m merely scratching the surface of an infinite number of questions around this topic. Please share with us your thoughts. Thank you.
– Timothy E.
Hi Timothy,
Great question about a topic that I think about all the time.
You may be surprised to know that machine learning, a form of AI, has been in use in the financial markets since the 1980s.
Of course, the application of machine learning is far more advanced now today than it was back then, and trading systems and speeds have evolved radically since then.
Anywhere between 60–75% of all trades in the U.S. markets are algorithmic trades made by computers largely based on machine learning.
We’re in that world already. But we must understand that the majority of these machine learning-driven trades are very short-term. They tend to look for short-term arbitrage opportunities.
Even if they’re making only a few cents per share off of a trade, they are doing it at scale and hundreds of times a day, which means they can still make a lot of money despite very small profits from each trade.
That’s why at Brownstone Research, our trading and investment strategies are never trying to compete with these very short-term machine learning algorithms. That’s a battle we would surely lose.
When we develop trading strategies, we do so in a way that gives us an advantage.
For example, my team and I have developed a neural network, a form of artificial intelligence called deep learning, that I call The Perceptron. We use it for pattern recognition in the cryptocurrency markets. But rather than programming it to look for short, intraday trading opportunities, we designed it to look for patterns that suggest highly probable large increases in the asset price over a 45–60-day window.
You’ll be hearing a lot more from me about The Perceptron in the weeks ahead.
By “seeing” these patterns, we can get in before the algorithms trade, hold on and wait for the move, and then get out when the signal changes.
Using different time horizons than hedge fund algorithms is a way for us to ensure that we’re beating the market and beating all those computers using machine learning.
The same is true for our buy-and-hold, longer-term investment research services. We spend countless hours researching companies and industries looking for competitive advantages, dislocations, inflection points, and companies that are undervalued and misunderstood by the market.
Oftentimes, these are things that a computer can’t “see,” as the information isn’t quantifiable yet.
Everyone hated Tesla in early 2018 when I recommended it. Look at the company now. The market thought NVIDIA was just a small gaming tech company in February 2016 when I recommended it. Look at NVIDIA now.
Wall Street hated Uber years ago because it was deeply unprofitable when I first recommended it, but our analysis showed that it could sell assets, raise capital, restructure, and become free cash flow positive. Look at Uber now.
Again, we can use time to our advantage as humans. We can stack the deck in our favor, even against Wall Street, by seeing things that they don’t yet see.
And my team and I can continue to develop our own AI at Brownstone Research using deep learning to give us an advantage in identifying great trading and investing opportunities.
There will be a lot more to come on that topic – we’ve been in research and development mode and have made some major breakthroughs in this space.
Hello Jeff,
In your Bleeding Edge edition dated 9/4, there are pictures of Gollum holding an Nvidia card. I’m just curious. Did you use AI to generate that picture?
– Cash G.
Hi Cash,
Speaking of leveraging AI at Brownstone Research… you bet it is!
In fact, we use generative AI every day for image generation. That image was created using Midjourney. It’s a prompt-based text-to-image AI.
It’s a lot of fun for my team to work with the technology. I actually insist we do. Using a large number of AI-powered tools keeps us efficient and competitive and keeps us on top of the latest developments in AI.
That’s it for today’s AMA. Thanks again for the questions. My team and I love hearing from you all.
Have a great weekend everyone.
Jeff
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.
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.
