Are you a —
Buyer pathSupplier path
Investment banking for AI hardware procurement.
What LII does
LII sits between companies that need AI compute at scale and the suppliers who can actually fulfil it. We take the requirement, verify both sides of the table, run a competitive sourcing process, and stay on the transaction through escrow and delivery.
We hold no inventory and manufacture nothing. That is deliberate — it means we have no stock to move and no reason to steer you toward one supplier over another. We are paid when a deal closes, on terms disclosed before you commit.
For suppliers, that means we arrive with a defined requirement and a verified buyer behind it. You quote against it or you don't. We don't ask for exclusivity and we don't ask you to publish inventory.
Why work with LII
01
Instead of emailing a dozen vendors and comparing a dozen incompatible quote formats, you give the specification to us. We normalise every response so unit price, lead time, condition and logistics line up side by side.
02
A lab racing a training run and a data centre building a five-year footprint are not solving the same problem. We ask which axis you're optimising for before we approach anyone.
03
Suppliers are checked for authorisation, provenance and capacity to fulfil before they are allowed to quote. The grey market in AI hardware is real; screening it out is most of the work.
04
The same person handles intake, sourcing, escrow terms and delivery coordination. Nothing gets handed to an account manager who wasn't in the original conversation.
01
We contact you when there is a defined specification, a quantity, a timeline and a buyer who has already been verified. No lists, no drip sequences, no speculative enquiries.
02
Entity, purchase authority, funding path and delivery destination are confirmed on our side first. You spend your time quoting, not qualifying.
03
Your stock, pricing and customer list stay yours. We never list your inventory publicly and we never present your availability as our own.
04
Funds and release conditions are agreed in writing before anything ships. You are not chasing payment after delivery, and the buyer is not sending wires into the dark.
Deal philosophy
You can usually have two of these cleanly and the third with effort. Telling us which one is non-negotiable changes who we approach and what we ask them for. Knowing which axis a buyer is anchored to tells you how to shape a quote. We pass that context on with every requirement we send you.
Racks live in weeks, not quarters. We prioritise suppliers with hardware physically in hand and a logistics path we can verify, and we accept a wider price band to get there.
The lowest defensible cost per unit of usable compute. We widen the supplier set, open the conversation to previous-generation and secondary-market hardware, and give lead time more room.
Provenance, authorisation and warranty above all. Fewer suppliers clear this bar, quotes take longer, and the hardware that arrives is exactly what the paperwork says it is.
How a deal runs
Nothing moves to the next step until the previous one is signed off. You know where the deal is at any point. You enter at step three, with the requirement already defined and the buyer already verified.
What you're building, what it has to do, and when it has to be live. Specification, quantity, delivery location, timeline, budget envelope, and which axis you're optimising for.
Before we approach a single supplier, we confirm the buyer is real: entity, purchase authority, funding path, delivery destination. Suppliers clear the same bar in reverse — authorisation, provenance, capacity to fulfil.
The requirement goes to the supplier network. Three to five quotes come back and are normalised into one comparison: unit price, lead time, condition, warranty, logistics, payment terms.
You choose the supplier. We tell you where each quote is strong and where it is thin — including when the cheapest one carries risk we would not take — but the counterparty is your decision.
Funds move through a trusted escrow process. Inspection windows and release conditions are agreed in writing before anything is packed, so neither side is exposed while hardware is in transit.
Hardware arrives at your site and is inspected against the agreed condition. LII is paid a commission of 0.5–5%, or a flat sourcing fee — set and disclosed before you commit, never adjusted after the fact.
Start here
Send the requirement — even a rough one. If it is something we can source, you'll hear back with what we need to open sourcing. If it isn't, we'll say so. Send your categories, capacity and authorisation documents. Once verified, you go into the network and we come to you when a requirement matches.
Sourcing directory
Six categories, indexed. This is the full scope of what LII sources — from a single accelerator to a fully integrated cluster with the power and cooling underneath it.
Availability, generation and condition are confirmed against live supplier capacity at the point of quote, which is why you will not find prices or stock counts here. Send a requirement and you get real numbers against it.
The compute itself. Current and previous generation, new and secondary market, across vendors.
Typical unit · 8–thousands of cards
The chassis the accelerators live in, air-cooled or liquid, per-node or integrated at rack scale.
Typical unit · 1–64 nodes
The fabric that decides whether a cluster scales or stalls. Usually the longest-lead item in a build.
Typical unit · per-rack or per-fabric
Sized to the training job, not the marketing sheet — checkpoint bandwidth first, capacity second.
Typical unit · per-tier, sized to cluster
Power and heat. The part of an AI build that most often turns out to be the real constraint.
Typical unit · per-rack or per-room
When the requirement is a cluster rather than a part number, and several vendors have to agree with each other.
Typical unit · whole build
What this page deliberately omits
Prices
AI hardware pricing moves with allocation, generation and volume. A number published here would be wrong by the time you read it. You get a live quote against your actual requirement instead.
Stock counts
We hold no inventory. Availability belongs to the supplier network and is confirmed at the moment of sourcing, not cached on a marketing page.
Supplier names
Our network is the asset, and the suppliers in it did not sign up to be a public directory. They are named to you when they quote on your requirement.
About
Buying AI hardware at scale still works the way private markets worked before anyone brokered them: opaque pricing, unverifiable counterparties, and outcomes that depend on who you happen to know.
LII exists to put a process in the middle of that. Verify both sides, run a real competitive bid, move the money through escrow, and get paid on the close. Investment banking for AI hardware procurement — the analogy is the whole business model, not a tagline.
How we operate
01
No requirement goes to the network until the buyer is verified, and no supplier quotes until their authorisation and capacity are checked. This is the step most brokers skip and the one that causes every bad outcome.
02
0.5–5%, or a flat sourcing fee, agreed before sourcing begins. We never take an undisclosed spread on top of a supplier's price.
03
We do not buy, hold or resell hardware on our own book. There is no shelf we need to clear, which is what makes a recommendation worth anything.
04
Inspection windows and release conditions in writing, before anything ships. Neither side carries the counterparty risk alone.
Founder
LII exists because sourcing AI infrastructure is still handled through fragmented networks, informal introductions, and unverified inventory claims. Diego built LII to bring a disciplined, verification-first process to how AI hardware actually changes hands.
Operating under LGA Ventures, LII applies the same standard to every deal: real suppliers, traceable hardware, competitive quotes, and a transaction structure that holds up when real money is on the line.
Every requirement is reviewed personally before sourcing begins. Buyers work directly with Diego from first contact through close.
Where we operate
LII opens in California, where the density of AI labs, data centres and buyers is highest, then expands across the United States and internationally as the supplier network deepens.
We work with AI startups, research labs, data centres, cloud providers, enterprises standing up internal AI infrastructure, and universities. Deals of ten or more accelerators are where the sourcing process earns its keep, but we will take a smaller requirement seriously if it is real.
LII is a division of LGA Ventures, currently onboarding buyers and suppliers in California.
Contact
Buyers submit a requirement. Suppliers register capacity. Both go to the same person and both get a real answer, including when the answer is no.
Every submission is reviewed for legitimacy before sourcing begins — entity, authority, and in the case of suppliers, authorisation and provenance. It is slower than a contact form that fires straight into a CRM, and it is the reason the network is worth being in.