Why Bittensor Subnets Should Trade Like Small-Cap AI Equities

Why aren't subnets treated as small decentralized AI companies when it comes to investing? What do subnets need to show to prove they're similar to small-cap AI equities?

The answer is twofold. Enterprise revenue, stated in financial statements that can be used to run a discounted cash flow analysis, will strengthen a subnet’s value as a real revenue-generating investment. Second is partnerships with reputable companies that endorse or use a subnet’s AI product or service. Once a subnet has both, the market has a real reason to price it like a business instead of a token.

Two Sources of Liquidity

The Bittensor ecosystem needs more liquidity, and that liquidity can come from two very different groups. The first is the trenches, the memecoin traders who want something with an actual product and an AI theme instead of another picture of a dog. The second is the institutional and retail investors who would normally put money into small-cap equities, capital that's never touched crypto at all.

Converting the memecoin traders should be the easier lift. They're already comfortable with crypto and already know how to size a position in novel technologies. So why haven't they come to Bittensor already?

Converting the small-cap equity crowd is harder, but more valuable. Those are people trading out of a Charles Schwab account who've never held a token. Getting them invested in Bittensor means genuinely fresh capital outside of traditional crypto capital. Right now the best shot is through Grayscale or Yuma-type products, and both are built for institutions, not retail. So how do we capture this retail crowd and get them interested in Bittensor subnets?

I'll answer both of these questions below and explain why, although memecoin traders are the low-hanging fruit for liquidity, they are not the source of capital that aligns with Bittensor subnets. On the other hand, small-cap equity investors are looking for revenue and margin expansion within the AI ecosystem, which aligns with the revenue-generating mission that Bittensor subnets have ingrained in their minds.

Subnets vs. Memecoins

Let's start with the group already in the crypto ecosystem. Subnets can be seen as more confident, product-producing memecoins, the kind that proliferated on Pump.fun in 2024. Pump.fun is just the application that lets anyone launch one of these tokens, where the token can be just about anything from an AI startup to a frog meme. 

Bittensor is the network that actually incentivizes real product-market fit for the AI applications and services behind them. It's clear that Bittensor is of a higher intellectual pedigree than Pump.fun and its memecoins, and it's directly connected to AI in a way a joke token never will be. So why aren't subnets pumping, and why isn't TAO pumping the way a hot memecoin does?

Part of the answer is that subnets do not have the same launch moment that memecoins have. As new subnets join, there should be more trading activity around their "IPO," but that is not usually the case. Pump.fun uniquely positioned itself where traders were able to see new launches on the app and bid as soon as they launched. Bittensor subnets do not have this kind of launchpad feel, which is intentional, to keep pump-and-dumps minimal and long-term value accrual the focus. 

Another piece of the puzzle is that memecoins have dedicated KOLs that will post about a token and its associated memes constantly for a short-term pump. In contrast, Bittensor subnets are focused on building actual products and aren't budgeting for KOLs to constantly post memes and updates.

Therefore, although memecoin traders are already associated with crypto, they are not the best source of liquidity for Bittensor, as subnets are focused on becoming revenue-generating businesses rather than speculative tokens.

Subnets vs. Small-Cap AI Equities

The second liquidity group is a different animal. This group consists of investors who sit inside a brokerage account and prefer the protections of investing in equities, where there may be a perception of less risk. These investors are focused on deploying capital into semiconductors, memory, and data center companies. Those who want exposure outside the S&P 500 look to the Russell 2000 and identify smaller companies in the AI stack.

That's the real comparison. A subnet is, structurally, a small-cap decentralized AI company, and it should trade like one of the small-cap AI names sitting in the Russell 2000. Both are revenue-generating businesses built around AI products. The difference is everything after the revenue. A Russell 2000 AI stock comes with investor protections, quarterly disclosures, and a listing in a normal brokerage account. A subnet offers none of that yet. So the real question is, what do we do about it?

A lot is being done to align on the implementation of investor protections. However, more needs to be done voluntarily by subnets to disclose financial statements and show recurring revenue from enterprise customers. At the end of the day, subnets are startup experiments running in live competitions every day, but the end goal is to turn that experiment into a cash-flowing business. Finding product-market fit is the primary result of the Bittensor economics and flywheel, and once a subnet has reached it, it's time to start business development. Subnets, which produce cheaper, more democratized AI products like inference, training, memory, and cybersecurity, should be attracting enterprise customers and realizing revenue in order to spark investor interest from small-cap retail equity investors.

The second area subnets need to develop is credibility from outside the ecosystem. That comes in three grades: a partnership, an endorsement, and, the only one that actually matters, a paying customer from a notable startup or a Fortune 500. We've already seen Targon reach the first grade, co-authoring a confidential compute research paper with Intel engineers, and the news traveled well beyond Bittensor. The third grade is where the repricing happens: AI startups valued in the billions publicly confirming they run inference on Engy, storage on Hippius, or serverless compute on Chutes will be monumental.

All that to say, one breakout subnet that has product-market fit, audited revenue from enterprise customers, and a product provided at a lower cost with equivalent performance will bring retail investors into the ecosystem.

One Breakout Subnet Until Nirvana

Bittensor exists to build decentralized versions of centralized AI products at lower cost with equivalent or better performance. Once this happens, equity investors will wake up to the fact that Bittensor's decentralized subnets are the next wave of value capture. What subnet will break out? I am watching the subnets that are addressing the bottlenecks in the AI stack. These are the opportunities for Bittensor, and three subnets are already providing services to take advantage of these bottleneck opportunities.

Engy (SN53) is fiercely attacking the inference bottleneck and serves frontier open models with cryptographic proof that the model you paid for is the model that ran, and the team ran Kimi K3, a 2.8 trillion-parameter model, on 80 RTX 5090 gaming cards at 20 tokens per second on day one. Lium (SN51) rents compute peer-to-peer at a fraction of centralized cloud pricing, no KYC, settled in crypto.

RedTeam, a real-time cybersecurity market, announced that its detection solution is now being tested inside five more enterprise platforms carrying a combined valuation above $22 billion, reaching more than 900 million accounts and ingesting over a billion transactions weekly.

We need to start pushing these subnets in front of enterprise businesses and labs rather than at investors, because once revenue is being realized and these subnets are able to document and show it, that is when price follows. One breakout subnet is all it takes to get Jensen and the whole AI ecosystem talking about Bittensor. 

Real Revenue, Real Customers, Real Partnerships

In my eyes, the entire purpose of Bittensor is to create decentralized, equally performant versions of the already centralized AI products that are also cheaper to use. There are bottlenecks in the AI ecosystem, and those are the opportunities Bittensor subnets are taking advantage of.

We must highlight these subnets more effectively, not to investors, but to potential customers and labs, since these are the users who'd actually pay to use them. These subnets need recurring revenue, and that revenue will bring investors by default, without anyone needing to pitch them.

If we can run a real DCF on a subnet's revenue and free cash flow using audited financial statements and disclosures, that's when real money starts to flow in, from either side of the liquidity equation. Revenue alone isn't the finish line either. What will it take for big tech companies to start using these subnets' products? We need more partnerships, the kind that prove a subnet's product holds up against a customer with other options.

Real revenue and real partnerships are what turn a subnet into a small-cap AI company the market can actually price. That's the trade Bittensor is primed to make.