everyone talks about AI models getting smarter, but there's another side of the AI race people don't talk about as much.
the cost of running them.
training models, serving requests, and running AI agents all need computing power. and a big part of that comes down to access to GPUs.
the thing is, GPU rental prices can change with demand, hardware availability, and competition for capacity.
so what if you could trade exposure to those prices?
that's where @injective's Super Intelligence Compute Index (SICI) comes in.
â–¸ so what exactly is SICI?
SICI is an onchain index tracking the hourly rental prices of five Nvidia GPUs, with each one carrying a fixed weight:
- H100: 40%
- B200: 25%
- H200: 15%
- A100: 10%
- RTX 4090: 10%
the index is built and priced by LO using GPU rental-price data from compute marketplaces.
the H100 carries the biggest weight, so changes in its rental price have more influence on the index than any other GPU in the basket.
and the RTX 4090 is worth noting. not every AI workload needs a massive data-center GPU. developers also use consumer cards for inference and open-source models, so including one gives the index exposure to a different part of the compute market.
â–¸ why bring this onchain?
GPU rental prices are an important part of AI infrastructure, but getting financial exposure to those prices hasn't been straightforward.
SICI introduces an onchain perpetual market tied to the index. it's live on Injective mainnet, with USDC as the margin asset and up to 10x leverage.
that doesn't mean you're buying or renting GPUs themselves. you're trading exposure to an index based on their rental prices.
imo, that's the part worth paying attention to.
AI infrastructure isn't just about who builds the best models. the cost and availability of compute matter too.
and now, there's an onchain market built around tracking that side of the industry.
Cleveland-Cliffs ($CLF) just gave the market a $4B reason to get bullish.
i'm not convinced that's the number investors should be focused on.
the real question is how much of that earnings power survives a weaker steel market.
for starters, the recovery is real.
- adjusted EBITDA more than doubled in Q3, and the company went from a loss to a profit. so yeah, there's actual improvement here, not just a bullish forecast.
- management expects over $4B in adjusted EBITDA for 2027. better customer contracts are a big part of that, potentially adding over $1B in year-over-year EBITDA before changes in steel prices.
and tbh, the contracts are the part i'd pay attention to. minimum-volume commitments can make earnings more predictable, but they don't magically remove the risks that come with being a steelmaker.
US tariffs also give domestic producers some protection from foreign competition. but if demand from automakers and other industries slows, or steel prices fall, CLF can still feel the hit.
imo, the real test is pretty simple:
can the company keep earnings strong and generate enough cash to reduce debt if steel prices stop rising?
if yes, there's a stronger case for a lasting turnaround. if profits take a hit as soon as the steel market cools, this might be more of a cycle play than the comeback investors are hoping for.
i wouldn't rush to chase the rally just yet. i'd want to see the forecast translate into actual earnings and cash flow first.
a good quarter is one thing. keeping it going when the market turns is another.
@officialmudrex
Disclosure: Mudrex offers US-stock perpetuals for users following US stocks from India. They provide price exposure, not ownership of the underlying shares, and carry their own risks.
