From NVIDIA to Binance, selling shovels is the strongest business model.

Written by: Liam, Deep Tide TechFlow

In 1849, during the California Gold Rush, countless people with dreams of getting rich flocked to the American West.

German immigrant Levi Strauss originally wanted to join the gold rush, but he keenly discovered another business opportunity: miners' pants often tore, and there was a pressing need for more durable work clothes.

Thus, he made a batch of denim jeans from canvas, specifically to sell to gold miners, which gave birth to a clothing empire called “Levi's.” However, the vast majority of those who actually participated in the gold rush lost all their money.

On November 20, 2024, NVIDIA delivered yet another “amazing” financial report.

Q3 revenue reached a record of 57 billion USD, a year-on-year increase of 62%; net profit was 31.9 billion USD, soaring 65% year-on-year. The latest generation of GPUs remains a scarce commodity that “even with money, one might not be able to buy,” and the entire AI industry is working for it.

At the same time, this script is also playing out in the cryptocurrency world on the cyber shore.

From the ICO bull market in 2017, to the DeFi summer in 2020, and then to the Bitcoin ETF and Meme wave in 2024, retail investors, project teams, and VCs continue to rotate through each narrative and each wave of wealth stories, but only exchanges like Binance consistently stand at the top of the food chain.

History always rhymes.

From the California Gold Rush of 1849 to the cryptocurrency frenzy and the wave of AI, the biggest winners are often not the “gold miners” directly competing, but those who provide them with “shovels”. Selling shovels is the strongest business model that crosses cycles and harvests uncertainty.

AI gold rush has made Nvidia rich.

In public perception, the protagonist of this wave of AI is undoubtedly the large models represented by ChatGPT, which are intelligent agents capable of writing copy, creating art, and coding.

But from a commercial and profit perspective, the essence of this wave of AI is not an “application explosion”, but an unprecedented computing power revolution.

Like the California Gold Rush of the 19th century, tech giants such as Meta, Google, and Alibaba are all prospectors, engaging in a gold rush for AI.

Meta recently announced an investment of up to $72 billion in artificial intelligence infrastructure this year, and stated that spending will be even higher next year. CEO Mark Zuckerberg stated that he would rather risk “missing out on hundreds of billions of dollars” than fall behind in the development of superintelligence.

Companies such as Amazon, Google, Microsoft, and OpenAI have made record capital expenditures in the field of AI.

The tech giants have gone crazy, and Jensen Huang can't stop smiling; he is the Claude Lévi-Strauss of the AI era.

Every company that wants to build large models needs to purchase GPUs on a large scale and rent GPU cloud services. Every iteration of the model consumes enormous amounts of training and inference resources.

The model can't compete with its rivals, and there's no clear path to commercialization for the application. It can be rebuilt from scratch, but the GPUs purchased and the computing power contracts signed have already been paid for in real money.

In other words, on the proposition of “Can AI change the world?” and “Can AI applications be profitable in the long term?”, everyone is still exploring. However, if you want to participate in this game, you must first pay an “entrance fee” to the computing power providers.

NVIDIA is right at the top of this computing power food chain.

It has almost monopolized the high-performance training chip market, with H100, H200, and B100 becoming the “golden shovels” that AI companies are competing for. It has integrated the software ecosystem (CUDA), development tools, and framework support from GPUs, further forming a dual moat of technology and ecosystem.

It does not need to bet on which big model will win, it only requires the entire industry to continuously “gamble”: betting that AI can create some kind of future and can support higher valuations and budgets.

In the traditional internet, Amazon's AWS used to play a similar role. Whether a startup can survive is one thing, but sorry, you first need to pay for the cloud resources.

Of course, NVIDIA is not isolated; behind it is a whole “shovel supply chain” that is also laughing all the way to the bank in the wave of AI.

GPUs require high-speed interconnection and optical modules. In the A-share market, Xinyise, Zhongji Xuchuang, and Tianfu Communication have become an indispensable part of the “shovel,” with stock prices rising several times this year.

The transformation of data centers requires a large number of cabinets, power systems, and cooling solutions. From liquid cooling and power distribution to the infrastructure of the computer room, new industrial opportunities are constantly emerging; storage, PCB, connectors, and packaging testing, all manufacturers of components related to “AI servers” are reaping valuations and profits in this wave.

This is the terrifying aspect of the shovel-selling model:

Gold diggers may lose money, and gold mining activities may fail, but as long as people are still digging, those who sell shovels will never lose.

The large models are still struggling with “how to make money,” while the computing power and hardware supply chain are already steadily counting profits.

Cryptocurrency miner

If NVIDIA is the shovel seller of AI, then who is the shovel seller of Crypto?

The answer everyone can think of: exchange.

The industry is always changing, the only constant is that the exchange keeps printing money.

The year 2017 was the first true global bull market in the history of cryptocurrency.

The threshold for issuing tokens in projects is extremely low; a white paper and a few PPTs can lead to financing and launch. Investors are crazily chasing “tenfold and hundredfold tokens,” countless tokens go online only to return to zero, and most projects are frozen or delisted within 1-2 years, with even the founding teams disappearing from the timeline.

However, there are fees for listing projects, users have to pay transaction fees, and futures contracts charge fees based on positions.

The price of the coin can be halved and halved again, but the exchange only cares about trading volume to make a profit; the more frequent the transactions, the more intense the fluctuations, and the more they earn.

In 2020, during the summer of DeFi, Uniswap challenged traditional order books with the AMM model, and various mining, lending, and liquidity pools made it feel “as if centralized exchanges were no longer needed.”

However, the reality is very nuanced, with large amounts of funds moving from CEX to on-chain mining, and then returning to CEX during peaks and crashes for risk management, cashing out, and hedging.

In terms of narrative, DeFi is the future, but CEX remains the preferred entry point for deposits, withdrawals, hedging, and perpetual contract trading.

By 2024-2025, Bitcoin ETFs, the Solana ecosystem, and Meme 2.0 will once again push crypto to new heights.

During this cycle, whether the narrative has turned into “institutional entry” or “on-chain paradise”, one fact remains unchanged: there is still a large amount of leveraged capital flowing into centralized exchanges; leverage, futures, options, perpetual contracts, and various structured products constitute the “profit moat” of the exchanges.

In addition, CEX is also integrating with DEX at the product level, making on-chain asset trading a norm in CEX.

The price of coins can rise and fall, projects can rotate, regulations can tighten, and sectors can shift, but as long as everyone is still trading, as long as volatility remains, exchanges are the most stable “shovel sellers” in this game.

Besides exchanges, there are many “shovel sellers” in the crypto world:

For example, mining machine companies like Bitmain profit by selling mining machines rather than mining, allowing them to remain profitable through multiple rounds of bull and bear markets.

Infura, Alchemy, and others provide API services and benefit from the growth of blockchain applications.

Issuers of stablecoins like Tether and Circle earn “seigniorage” on digital dollars through interest rate spreads and asset allocation;

Pump.Fun and other asset issuance platforms continuously extract taxes through the bulk issuance of Meme assets.

……

In these positions, they don't need to bet on which chain will win or which Meme will explode each time; as long as speculation and liquidity are still there, they can print money stably.

Why is “selling shovels” the best business model?

The real business world is far more brutal than people imagine; innovation often comes with great risks. To succeed, one not only needs personal effort but also relies on the course of history.

In any cyclical industry, the results are often like this:

Building upper-layer applications is like mining for gold, pursuing Alpha (excess returns). You need to bet on the right direction, bet on the right timing, and outsmart your opponents. The win rate is extremely low, and the odds are very high; a slight misjudgment could lead to total loss.

Building the underlying infrastructure is like being an upstream shovel seller, earning Beta. As long as the entire industry continues to grow and the number of players keeps increasing, one can reap the benefits of scale and network effects. Shovel sellers are in the business of probability, not luck.

NVIDIA does not need to choose which AI large model can “run out”, and Binance does not need to determine which narrative can last the longest.

They only need one condition: “Everyone continues to play this game.”

Moreover, once you get used to the NVIDIA CUDA ecosystem, the cost of migration is unimaginable. Once your assets are all on a major exchange and you are accustomed to its depth and liquidity, it becomes very difficult to adapt to a smaller exchange.

The ultimate outcome of the shovel-selling business is often monopoly. Once a monopoly is formed, the pricing power is completely in the hands of the shovel sellers, as evidenced by Nvidia's gross margin of up to 73%.

To summarize from a very blunt perspective:

Companies that sell shovels earn the “tax of industry existence,” while companies that mine for gold earn the “time window dividend.” You must seize the user's mindset during a brief window of opportunity, or you will be abandoned. Those who create content or narratives earn “money from attention fluctuations,” and once the trend shifts, traffic evaporates immediately.

To put it more bluntly:

Selling shovels is betting that “this era will go down this path”;

Building an application is betting that “everyone will only choose my company.”

The former is a macro proposition, while the latter is a brutal elimination match. Therefore, in terms of probability theory, the win rate of selling shovels should be an order of magnitude higher.

For us retail investors or entrepreneurs, this is also a profound revelation: If you can't see who the final winner is, or don't know which asset will continue to rise several times, then invest in the person who provides water to all the miners, sells shovels, or even just sells jeans.

Finally, let me share one more piece of data: Ctrip's Q3 net profit is 19.919 billion, surpassing Moutai (19.2 billion) and Xiaomi (11.3 billion).

Don't just focus on who shines the most in the story,

Think about who can continuously charge in all stories.

In a fervent era, serving the fervor while maintaining calm is the highest wisdom in business.