Jensen Huang and Michael Dell discussed the supply bottleneck: Michael - memory is the challenge. Jensen - we will have billions of agents, and we will need to give each agent essentially its own computer to use.

Watch full here:

Bullish for $MU $DRAM $PENG 🤩

Crank it up $MU

The memory problem isn’t going to be solved by scaling hardware

But memory has topped so I guess they’ve solved the bottleneck 😂😂😂

HBM patents stolen from NETLIST (NLST) by MU , GOOGL SAMSUNG & SK? Jensen Netlist is in Irvine! @nvidia @MichaelDell

Jensen will soon lose the huge respect he had by constantly pumping NVDA He doesn’t have to stoop low JMO

TrendForce data backs this: mobile DRAM contract prices still rising in 2Q26, squeezing smartphone production. Memory is the binding constraint, not compute.

@EdLudlow Ed, You should do an interview with CEO OF Netlist ?

If every agent effectively needs dedicated compute resources, the long-term demand for servers, networking, memory, and power could be much larger than people expect.

agents need crazy resources. memory's the bottleneck.

Every agent its own computer. Massive baseload demand. Wind and solar fail at this scale. Nuclear keeps the lights on.

10/10 way to sell more servers

Io net

Check

Supply bottlenecks often result in massive shifts in pricing power.

Every agent gets a computer, storage, and long-term memory. The right infrastructure vision. The next layer: grounding that memory in your business data. That turns compute into context - and agents into teammates that actually know your work.

AI agents become mainstream, this won’t look like a normal tech cycle anymore. It starts looking like the internet buildout, cloud revolution, and industrial electrification all happening at once.

Man, people dont know whats coming. Memory stocks have & will make many more millionaires.

Two guys selling product

Dell sees the hardware ceiling. Jensen sees the civilization scale demand above it — both are right and that gap is where the next decade of investing lives. Billions of agents each needing dedicated compute isn't a product roadmap — it's a complete reimagining of how much infrastructure the world actually needs. Memory bottleneck today, power bottleneck tomorrow, and somewhere in between an opportunity most investors haven't priced yet.

$MRAM

@Theta_Network will have many homes for agent work at the edge.

“Each agent needs its own computer” is a hardware sales thesis, not an architectural truth. Stateless, intent-native Aptivs share compute. The memory bottleneck @MichaelDell named is structural to LLMs, not to all AI. There’s another way.

Giving every agent its own computer is one approach. Teaching infrastructure to wake instantly, run, and disappear is another.

That’s what stands for