I smelled it before I read it. A press release on Crypto Briefing—a media outlet built on ICO hype and exchange listings—announcing a $55 million Series A for a "vision AI startup" founded by ex-DeepMind researchers. No GitHub. No whitepaper. No testnet. Just a promise to "redefine industry standards." My on-chain alarm bells clanged harder than a flash loan attack.
Let me be clear: I don't trust press releases. I trust block explorers, transaction hashes, and raw code. And when a story about a supposedly revolutionary AI company breaks on a crypto-native outlet, my first instinct is to trace the money—not the narrative.
I pulled up Etherscan. Polygon. Solana. Zero contracts. Zero token deployments from any entity matching "Elorian." I checked the team's LinkedIn. Two ex-DeepMind researchers, yes—but their last publications were in 2022 on agent-based reinforcement learning, not vision. The company has no public API, no demo, not even a blog post explaining their approach. For a $300 million valuation, that's a red flag the size of a CryptoPunk.
Context: The Perfect VC Narrative
Elorian is a London-based startup founded in early 2024 by Dr. Alice Chen and Dr. Mark Liu, both former research scientists at DeepMind's structured reasoning team. The company claims to be building a "novel vision AI architecture" that outperforms existing models on complex scene understanding by 10x in accuracy and 5x in inference speed. Their pitch deck—which I obtained through a backchannel—includes vague references to "state-space models optimized for visual tokens" and a "proprietary training data pipeline."
But the deck lacks any benchmark numbers. No ImageNet results. No COCO scores. No comparison to SAM, DALL-E, or GPT-4 Vision. The only concrete claim is a $55 million Series A led by a previously unknown fund called "Nexus Capital Partners," with participation from a few crypto-native VCs I'd never heard of—Blockchain Ventures and Signal Fund. The valuation: $300 million post-money, implying ~18% dilution.
In the current AI funding climate, a $300 million valuation for a pre-product, pre-revenue team with elite credentials isn't unheard of. Stability AI hit $1 billion after releasing Stable Diffusion. But those cases had tangible outputs. Elorian has nothing.
Core: Seven Dimensions of Empty Hype
I ran Elorian through my standard due diligence framework—the same one I used in 2020 to spot the Curve Finance audit delay and in 2021 to identify NFT metadata scams. Here's what I found, or rather, didn't find.
1. Technical Route: A Black Box
Without a single line of code or paper, I can only infer their approach from the founders' backgrounds. Chen and Liu published papers on hierarchical reinforcement learning and multi-agent systems—not vision. Their last relevant work was a 2022 paper on "Visual Reasoning in Grid Worlds" that used symbolic reasoning modules, not deep learning.
If Elorian is indeed pursuing a state-space model (SSM) for vision—like Mamba but adapted for pixels—they'd be entering a crowded field. Stanford's S4, together with Princeton's Mamba, have already shown SSMs can rival transformers on long-sequence tasks, but their application to image/video is still experimental. The claim of 10x accuracy over SAM is absurd without evidence. I asked a former Google Brain researcher (who wishes to remain anonymous) about the feasibility: "If they had such a breakthrough, they'd publish immediately. Not hiding behind a press release on Crypto Briefing."
2. Commercialization: The $55M Mirage
$55 million at a $300M valuation means investors paid for potential, not performance. In my 2024 interview with a BlackRock ops manager, I learned how institutional capital flows into AI: they demand proof-of-concept and multi-year contracts. Crypto VCs are different—they bet on teams and narratives, often with token economics in mind.
Elorian has zero revenue. Zero customers. Their business model slides mention "enterprise API" and "vertical SaaS for autonomous driving," but no target market size or go-to-market plan. The CFO (an ex-banker from Goldman) told me in a brief LinkedIn message that they plan to "start monetization within 12 months." That's typical runway math: $55M burns at ~$20M/year for a 50-person team, leaving 2.5 years. But GPUs for training a 100B-parameter vision model cost $10M per month if using H100 clusters. Either they have a massive cloud credit deal, or the math doesn't add up.
3. Industry Impact: None Yet
Assuming their tech works, it could disrupt video surveillance, autonomous robots, and medical imaging. But those industries are already served by established players like C3.ai, Scale AI, and countless startups. The barrier to entry from an unknown entity is enormous. I saw this in 2021 when I scraped metadata of 500 NFT collections: 15% had broken links. The hype-to-reality ratio here is similar.
4. Competitive Landscape: David vs. Goliath
Competitors include DeepMind itself (the founders' former employer), OpenAI (GPT-4 Vision), Meta (SAM 2), and NVIDIA (Cosmos). Each has orders of magnitude more compute, data, and talent. Elorian's only differentiator is a claimed "novel architecture." Without proof, they're a startup with a story, not a technological moat.
5. Ethics & Safety: Ignored
The press release mentions no bias testing, no privacy measures, no content moderation. Visual AI is particularly prone to demographic biases and misuse. For a company claiming to build "trustworthy AI," the silence is deafening.
6. Valuation: Crypto-Fueled FOMO
A $300M valuation for a Series A is high, but not unprecedented. The concern is the investor mix: Nexus Capital Partners has no track record in AI; Blockchain Ventures and Signal Fund are pure crypto VCs. This suggests the real exit might be a token sale rather than an acquisition by Google. I traced Nexus's wallet on Ethereum—they deployed a smart contract for a previous portfolio company that later launched a governance token. The pattern is clear: use AI hype to attract capital, then pivot to a token-based revenue model. Elorian's CEO, Dr. Chen, declined to answer whether they plan to issue a token. That's a non-answer that speaks volumes.
7. Infrastructure: GPU Gravy Train
Training a modern vision model requires thousands of H100s. $55M barely covers six months of a large cluster. Elorian claims they have a "strategic partnership" with a cloud provider, but won't name which. My investigation found no AWS or GCP credit announcements. If they're relying on decentralized compute (like Render Network or Akash), that would explain the crypto media outlet—and would also explain why their technical claims are so thin. Decentralized GPU is still orders of magnitude slower and less reliable than centralized solutions for training large models.
Contrarian Angle: The Crypto Briefing Signal
Why would a legitimate AI startup choose Crypto Briefing for its funding announcement? Traditional outlets like TechCrunch, VentureBeat, or even The Block would provide broader reach and credibility. Crypto Briefing has a history of publishing paid press releases for projects that later turned out to be scams or vaporware. I checked their previous coverage: in 2022, they ran a similar story for a "metaverse AI" startup that raised $30M—the company folded within a year after failing to deliver a product.
This suggests Elorian's target audience isn't enterprise customers or developers. It's crypto investors who might buy into a future token offering. The entire exercise could be a marketing campaign to drive up expectations for an NFT or token sale. I've seen this before: in 2021, a project called "Fetch.ai" raised millions on AI buzz, then took years to ship. Elorian is following the same playbook.
Furthermore, the lack of any on-chain activity from the claimed investors raises questions. I tried to validate the $55M transfer on-chain. No evidence. The traditional VC world uses wire transfers, but crypto VCs often use stablecoins. I found no USDC or USDT flows matching the amount. Either the deal closed off-chain (normal), or the announcement is purely speculative.
Takeaway: Wait for the White Paper, Not the Press Release
My track record as a news cheetah is built on on-chain verification. I broke the CryptoKitties gas crisis by reading the smart contract, not the press releases. I exposed the Curve audit delay by testing the protocol myself. Elorian has given me nothing to verify. Until they publish a technical paper with benchmarks, deploy a working demo, or reveal their investor wallet addresses, this announcement is no different from the countless DeFi past: shiny numbers, empty substance.
If you're an investor, don't chase the narrative. Wait for the data. If you're a builder, watch for their GitHub. I'll be monitoring for any token deployments under the Elorian name. When the smoke clears, we'll see whether this was a genuine AI breakthrough or another crypto-funded illusion. My bet is on the latter—and I'm putting my money where my mouth is: I shorted the Ethereum name for "Elorian" before it could trend.
Still want to invest? Send me the whitepaper. Until then, I'm running my Python scripts to scrape any mention of "Elorian" on-chain. Because in a market that moves sideways, the only news that matters is the news you can confirm with a hash.