Daniel Lai

Global Co-Head of Al/Compute Infrastructure @ Cantor Group

[brexton @brexton]

The scarcest and most in-demand resource on earth right now is GPU financing for sub-IG/unrated offtakers and ASICs

This tweet will appeal to approximately 6 people

[Shanu Mathew @ShanuMathew93]

Useful @EpochAIResearch piece on how increasingly complex AI financing structures actually work. The Anthropic deals are a good example:

>Compute: SPV buys >1 GW of TPUs → leases to Anthropic → institutional capital funds ~$35B → Broadcom backstops most senior debt

>Data centers: project SPVs fund construction → Fluidstack leases capacity → Anthropic uses it → Google backstops portions of the rent

The stack increasingly looks like: AI demand → long-term contract → SPV/project company → institutional capital → vendor/hyperscaler credit support

Where I disagree slightly is the conclusion that financing therefore isn't a constraint.

>These deals show you can finance incremental 1–2 GW projects. But we're talking about potentially mid-teens GW of incremental capacity in North America and ~30 GW globally. That quickly becomes hundreds of billions to >$1T of capital!

At that scale, the marginal decision-maker influences outcomes and the controlling party is shifting. It is no longer just convincing hyperscaler C-suites that AI has enormous long-term value. You increasingly have to convince banks, insurers, bond funds and the marginal syndicated lender that these cash flows are durable enough to repay them.

So, I would frame it less as “financing isn't a constraint” and more as financial engineering has pushed the constraint outward.

We also don't know a world where these SPVs blow up and hyperscale/OEMs get enforced. Have a feeling these guys negotiated much more leverage than they let on while promoting these vehicles...

[Daniel Lai @daniel_lai]

Major moves @Maneesh_Apte

[maneesh @Maneesh_Apte]

For decades the best ad creative knowledge lived inside a handful of top ad agencies.

Today we open-sourced it.

If you want to increase high quality creative volume and unlock AI powered growth all within your existing Claude setup, this is for you.

Comment for access, limited spots available.

This is the start of the era of agentic marketing!!

[Daniel Lai @daniel_lai]

@Maneesh_Apte Love this

[Daniel Lai @daniel_lai]

@kushalbyatnal Congrats!

[Daniel Lai @daniel_lai]

Scaling GPU financing to non-IG offtakers will be the next frontier

The long tail of compute demand doesn’t live with the hyperscalers

[Daniel Lai @daniel_lai]

For AI data centers, time to power is really time to gas, and time to gas is a question of pipelines.

Reserving capacity requires 15-20 year commitments most data center developers can't post.

Securing balance sheet backing will be critical to lock down pipeline capacity.

[brexton @brexton]

Tech twitter I need your help

What insurance startups/companies/products are there that protect lenders against GPU residual value risk

Please DM or tag if you know anyone (or if you’re a founder interested in working on this)

[Daniel Lai @daniel_lai]

When LLMs and agents become commoditized, malicious network attacks will accelerate and scale significantly

Security will soon require vertically integrated compute to make the cost of defense financially sustainable

[Daniel Lai @daniel_lai]

@jslishi IMO we’ll see some form of UBI effectively get implemented from the ground up by developers and tenants.

[Daniel Lai @daniel_lai]

We're going to see a massive increase in community benefit agreements become a significant cost for AI tenants.

Leverage has completely switched from municipalities competing for developers with tax incentives to pressuring developers for sweeteners to get anything built

[Axios @axios]

Sanders and AOC unveil data center moratorium bill https://t.co/83zQ4r79AB

[Daniel Lai @daniel_lai]

@Melt_Dem @brexton Love this, thanks for sharing!

[Daniel Lai @daniel_lai]

I'm excited about next-gen dielectric fluids not only for increasing server densities, but also for helping expand the useful life of components (reduced oxidation and corrosion and eliminating micro-vibrations from server fans).

Interesting to see which fluids take off here

[Daniel Lai @daniel_lai]

@jslishi Thanks for the note!

[Daniel Lai @daniel_lai]

The combination of a skilled labor shortage alongside fast advancing data center infrastructure is the perfect combination for more outages caused by human error.

Agentic DCIM could be a solution but how many data centers have sufficient telemetry to even try this?

[Daniel Lai @daniel_lai]

Inter-server networking today accounts for ~8% of AI data center power.

Great to see new advancements like MicroLED drive down these costs.

[Daniel Lai @daniel_lai]

That was quick

[DCD @dcdnews]

Akash Systems launches AMD-powered diamond cooled AI servers https://t.co/2HGOf5Bpag

[Daniel Lai @daniel_lai]

Thermal capacity and heat dissipation will become the next frontier for AI chips and racks as never ending inference and training push chips to their maximum power draw.

New substrates like diamond CVD will arise as the benefits of higher performance and component life outweigh the costs.

[Daniel Lai @daniel_lai]

The data centers being built today for LLM training & inference are mismatched for a future of world models

Developers who truly want to future-proof their builds should be focused on enabling real-time inference and storage ready for massive volumes of video and sensor data

[Daniel Lai @daniel_lai]

Sustainability is clearly a low priority but I’m excited how Finland is integrating their data centers in district heating.

Historically air-cooled data centers produced low-grade heat but AI liquid cooling racks are changing the economics of repurposing the heat.

[Daniel Lai @daniel_lai]

While I believe immersion cooling will soon be a necessity as rack density increases, direct-to-chip liquid cooling has a lot of room to innovate and is MUCH easier to retrofit