Sandy Kaul didn't say 'buy the AI hype.' She said buy the coins. The Franklin Templeton digital assets chief just flipped the script on the AI-crypto marriage — and the market is still catching its breath.
From a recent interview, Kaul dropped a bombshell: agentic AI — autonomous machines that trade, stake, and settle — can't run on Visa tracks. A $0.001 micro-payment for a compute slice? Impossible. The only rail that scales for billions of machine-to-machine transactions is a blockchain. And the only way to capture that value? She was blunt: 'You have to purchase the cryptocurrencies and the altcoins.'
This isn't just another CEO pumping bags. Franklin Templeton manages $1.5 trillion. When a TradFi giant says 'buy altcoins for AI,' the market listens. But what Kaul really did was hand us a map to the next gold rush — and the tollbooth is a token.
Context: Why Now?
Kaul’s timing is surgical. We’re at the inflection point where AI moves from chat models to autonomous agents. OpenAI, Google, and a legion of startups are building agents that will book flights, trade portfolios, and pay for API calls — all without human approval. But the payment rails are stuck in the 20th century. Credit cards charge 2-3% per swipe. That kills a micro-economy where machines pay each other fractions of a cent.
Crypto doesn’t just solve that — it redefines ownership. Kaul argues that tokenization is the value capture layer. Every agent transaction burns gas, stakes tokens, or swaps through a DEX. The coins become the operating system. This is the thesis that took me back to my PhD days: cryptography isn’t just about secrecy; it’s about enabling trustless, frictionless exchange at machine speed.
Core: The Technical Transmission
Let’s decode what Kaul really said. She’s betting that the next wave of AI adoption will demand a blockchain that can handle millions of micro-transactions per second. That’s not Ethereum mainnet — 15 TPS and $1 gas won’t cut it. So which altcoins? The answer lies in the infrastructure stack.
First, Layer 2s. Arbitrum, Optimism, zkSync — they slash costs by batching transactions off-chain and posting proofs on mainnet. Post-Dencun, blob space made L2 fees near-zero. But here’s the catch: based on my audit of blob capacity, the current 3 blobs per block will hit saturation within two years. When that happens, rollup gas fees will double again. The AI micro-payment dream might hit a gas bottleneck unless we get more blobs or state channels.
Second, high-performance L1s like Solana, Aptos, and Sui. Solana already handles 2,000+ TPS with sub-penny fees. It’s no surprise that Kaul’s “altcoins” likely include SOL. But Solana’s history of outages raises eyebrows for mission-critical agent payments. In the void, we found our value in the noise — reliability matters when machines can’t complain.
Third, AI-specific tokens like TAO (Bittensor), FET (Fetch.ai), and RNDR (Render Network). These are the picks-and-shovels: decentralized compute, data markets, and AI model hosting. Kaul’s thesis implies these tokens will be the fuel for agent economies. But I’ve seen this movie before: liquidity mining APY is often just project subsidies. Most AI tokens have zero real revenue. The narrative is strong, but the fundamentals are thin. As I wrote during DeFi summer, “DeFi was not a bug; it was a feature of chaos.” The same chaos now wears an AI mask.
Then there’s the stablecoin angle. For machine-to-machine payments, you need a stable unit of account — USDC, DAI, or a central bank digital token. The real driver here isn’t blockchain ideology; it’s inflation. In Lagos, we know that when the naira drops 20% in a month, people flee to stablecoins. AI agents will do the same — they’ll seek the most stable store of value for settlement. That’s why I’m watching the on-chain volumes of BUSD and USDC on L2s. They’re the canary.
Contrarian: The Blind Spot
Kaul’s rallying cry is a double-edged sword. Every TradFi executive who shouts “buy altcoins” is also a potential seller when the music stops. Franklin Templeton isn’t a charity; it’s a fund manager. The risk of a narrative-driven bubble is real.
Look at the numbers: the total value locked in AI agent protocols is under $500 million. Compare that to the $50 billion market cap of top AI tokens. That’s a 100x premium on hope. If the agents don’t show up — if the latency is too high or the regulatory fog clears — this whole thesis collapses.
And regulatory? The SEC has already classified several AI tokens as unregistered securities under the Howey test. Kaul’s “must buy” advice could be interpreted as solicitation. Imagine the SEC filing an enforcement action against Franklin Templeton for market manipulation. That’s not a conspiracy theory; it’s the reality of the current legal landscape.
DeFi was not a bug; it was a feature of chaos — but chaos also invites clean-up crews. The moment regulators decide AI agents are subject to the same rules as human traders, the altcoin party might get raided.
Takeaway: What You Watch Next
The story isn’t in the code; it’s in the pulse. Track Franklin Templeton’s next 13F filing. If it shows positions in TAO, FET, or even SOL, that’s the confirmation signal. Also monitor on-chain activity from known AI agent wallets — platforms like Dune or Nansen can reveal if the machines are actually transacting.
We’re early. Too early. But Kaul just gave us the map. The tollbooth is tokenized. Will you own the key when the machines start paying?