Quantum Blockchain Technologies has a whizzy name and a complicated history.
Back in 2000, it was an internet incubator called Brainspark. About half its investments had failed within a year of its Aim float. An Italian entrepreneur, Francesco Gardin, then picked up the pieces via his listed software company. By 2010, Brainspark’s main asset was a plot of land an hour from Milan that Gardin hoped to turn into a theme park. Next came a water park an hour from Turin, some Milanese sushi restaurants and an Italian hotel chain that collapsed, by which time Brainspark was called Clear Leisure.
Luke Johnson chaired Clear Leisure for a year, resigning in October 2013 after financial irregularities at its African hotels subsidiary came to light. “Unfortunately the figures were nonsense and the business was infested with litigation and fraud,” Johnson wrote at the time .
Blockchain was added to the business plan in 2018, then in 2021 to the corporate name. Quantum Blockchain Technologies, still run by Gardin, has since wrapped its collection of distressed legacy investments and legal claims in crypto jargon.
In essence, Quantum Blockchain Technologies says it has discovered more efficient ways to mine bitcoin. The proof-of-work lottery that rewards miners for maintaining the blockchain has repeating patterns that make it possible to eliminate redundant computations, it says.
We asked leading academics in the field of crypto for their thoughts on this proposition. Many were sceptical.
Making any sense of the Quantum Blockchain’s claims requires a foundational knowledge of crypto, so what follows might be a test of patience. If you’ve done the 11 hours of study needed to become a Certified Cryptocurrency Expert™ , or if are happy to suffice with the “ solved Sudokus ” analogy, please skip forwards until we stop talking about balloons and teacups.
Bitcoin’s blockchain updates its ledger of transactions by creating a new block about every 10 minutes. Miners compete by making candidate blocks. Blocks always contain the digital fingerprint of the previous block, lots of transaction data, and an address to deliver the reward. The prize for having your candidate added to the blockchain is some newly-minted bitcoins plus any fees attached to the transactions being verified in the block.
Mining means generating a numeric fingerprint for a candidate block, known as a hash. If the hash happens to be a number that’s the same or lower than the target number set by the bitcoin protocol, that miner’s candidate block becomes the official one and the process starts over again.
An algorithm known as SHA-256 creates the hash. It takes all the data in the block, stirs it around in a random yet repeatable way and spits out a number. This signature number can be as high as a quattuorvigintillion, approximately. The same data will always generate exactly the same hash, but source data can’t be extracted from the hash. Changing a single bit in the block will change its signature in ways that are effectively impossible to predict.
To maximise this unpredictability, there’s a section of the block that’s just for doodling. The number that goes here is called the nonce, which is an old English word meaning single-use. It’s included because transaction data in the block will be very similar between miners, if not identical, but the nonce is literally any number. A proof-of-work lottery involves randomly generating new signatures by keeping the candidate block the same and changing the nonce, usually by some quick and easy method such as adding 1. This tiny change will completely rewrite the hash specific to that block, and where this rewrite will land among the approximately quattuorvigintillion permutations can’t be predicted. The only way to find out is to do the work.
Bitcoin’s protocol adjusts its target number to regulate how often new blocks are added. This number, known as the target difficulty, is a fixed length and starts with a lot of zeros. If the target difficulty has 10 leading zeros and a hash for a candidate block just happens to have a signature with 11 leading zeros, that block becomes the official one and its miner wins the prize. Hash generation is unpredictable enough for this process to be considered a chance lottery.
OK, that’s still a lot of jargon, but maybe we can try another analogy?
Imagine a water balloon hitting a target that’s spinning rapidly at a constant rate, suspended above a teacup. Or, if you’d prefer, look at the post illustration. Now imagine this happening in perfect laboratory conditions, meaning an identical throw of an identical balloon will result in an identical explosion.
Now imagine a game where the challenge is to get as little water as possible in the teacup. Contestants have nearly-but-not-quite-identical balloons and can only change the angle of their throw.
Realistically, contestants have absolutely no way of predicting how much water will end up in the teacup. Some might claim to have strategies but the only good option is to keep throwing slightly different balloons at slightly different angles until the right amount of water lands in the teacup. Once that happens the winner is declared and the game starts again, except this time everyone’s balloons are filled with gravy.
By knowing the exact balloon structure and the exact angle, a person can reproduce the winning throw. But importantly, it’s not possible to work out anything useful from the amount of liquid that lands in the teacup. Each result is random yet replicable. There’s no secret formula. Knowing how the water-throwing round went offers contestants no advantage whatsoever in the gravy-throwing round.
In this analogy, the balloon is a miner’s candidate blockchain block. The spinning target is the hashing algorithm. The teacup represents the difficulty level, where the maximum waterline is the target difficulty. The angle of throw is the nonce.
Congratulations, you’re now an Alphaville Certified Crypto Expert™.
Lots of people have tried lots of ways to optimise crypto mining. One of the most important is AsicBoost, which according to its inventors Timo Hanke and Sergio Lerner could lower bitcoin’s mining costs by about 20 per cent .
The basic idea of AsicBoost is to whittle down the range of potential guesses by splitting the hash calculation in two and reusing some data. It’s all hugely complicated but, in target-teacup terms, can be thought of a little like partially freezing some of the balloon.
Speculation that miners including Bitmain of China were covertly using this method or something similar caused an almighty rift in the crypto community and, in 2017, bitcoin changed how it recorded transactions in a way that made AsicBoost much less useful.
Quantum Blockchain says it has found several ways to reinvent AsicBoost and is seeking two UK patents. The first is for Asic Ultraboost, which it describes as using a similar two-part calculation method to its predecessor. The second is for ASIC EnhancedBoost, which involves “partial pre-computing of future blocks”.
In addition, Quantum Blockchain says it has developed three separate mechanisms for optimising mining.
Method A exploits a claimed weakness in block generation. Method B uses “a composition of ML and mathematical methods” to decide which blocks are most likely to win. Method C, announced earlier this month , does the same but with “predictive AI”. All of them, in one way or another, involve taking some of the random chance out of nonce selection.
Quantum Blockchain has yet to demonstrate its innovations publicly and publishes only outline information on how they might work. For Methods A and B the company provides short, typo-strewn YouTube videos that simulate how they might outperform conventional mining, followed by a disclaimer warning that all product images and descriptions are for illustrative purposes only:
Quantum Blockchain told FTAV it has not published any papers on technology or methodology “in order to protect their research”. Its patent applications are “as far as they have dared to go with making the developments public,” said a spokesman.
That policy leaves a lot of questions unanswered.
“Without further details, it is hard to say anything really useful about this,” Edgar Weippl , professor of security and privacy at the University of Vienna and a team leader at SBA Research , told FTAV:
In general terms, I would think it’s possible to not calculate the entire hash to see if it starts with the required numbers of 0 for bitcoin but to break early and thus save time/energy. I would imagine that this can be done on a hardware level [ . . . ]. However, this would then only apply to bitcoin and very similar cryptos. and only [until] they decide to change the hash algorithm.
A big challenge for Quantum Blockchain is that its claims go against the academic orthodoxy that SHA-256 and related algorithms are unbreakable using currently available computing power. Trust in the hash makes them ubiquitous tools for guaranteeing online security. If bitcoin’s implementation of SHA-256 has been churning out hashes in predictable patterns it could have far-reaching implications that would touch nearly every aspect of our digital lives.
“Generally, the security of hash functions is well-studied in the cryptography community,” said Zhuo Cai , of Hong Kong University of Science and Technology, who has published several papers on randomness in decentralised networks. “The reason that we are still using hash functions like SHA-256 is that no one has found a significant attack even though many smart people have tried a lot in the last 20 years.”
Quantum Blockchain’s announcement that it can find these patterns with predictive AI also raised some doubts. The gist of the company’s claim is that AI can sense before the full hash is generated whether a calculation will fall within the likely range of success. Terminating losers early improved mining efficiency by around 30 per cent, it says.
“I don’t think their approach is feasible,” Zhuo Cai told FTAV. Machine learning is good at solving structural problems such as face recognition, and for taking a sequence and approximating what comes next, whereas data produced by hash functions is unstructured and discontinuous by design, he said:
It is ridiculous that they claim that they can predict [the hash] for future blocks. If this is possible, then it infers that machine learning can predict all transactions in the next hour in the future. Note that even if the prediction is different than the reality by very little bit, the hash function outputs are vastly different.
Who’s behind the technology? Quantum Blockchain says on its website that its R&D team “is composed of a number of sector experts selected across the UK and Italy, and includes highly skilled professionals, PhD students and university professors, with expertise in Quantum Computing, Machine Learning, Cryptography and Algorithms Optimisation Theory.”
The company’s 2022 annual report registers just four employees, up by one on the previous year, which includes its three directors. The 2023 edition is due next quarter.
“Team members [have] asked to remain confidential as one team member has been approached by someone from the industry,” Quantum Blockchain told us. “All annoyingly hush hush until they commercialise the product. So, everything publicly available is on RNS or the website.”
On LinkedIn there are two profiles giving Quantum Blockchain as their current employer: chief research officer Rita Pizzi , a hire it announced in November 2022 , and “data scientist freelance” Isacco Valsecchi , both of Lombardy in Italy.
Going by their CVs, Valsecchi graduated in 2021 with a bachelors degree in musical informatics from the University of Milan’s computer science department, where Pizzi was a senior research fellow between 2002 and 2022 . Describing his “work experience” at Quantum Blockchain, Valsecchi says he has “reverse-engineered bitcoin mining protocols” and “regularly communicated complex technical concepts to non-technical stakeholders”.
None of these peculiarities are out of character for Quantum Blockchain, which even by the wild-west standards of London’s junior market has been an odd investment. Its annual revenue has been zero or negligible in 21 of the 23 years since its foundation. Its aggregate net loss over the whole period, with 2023 figures not yet available, has exceeded £100mn:
Quantum Blockchain’s main backer for at least a decade has been Eufingest, a Swiss-based company that has provided several loans and to our knowledge has never declared another investment.
Since announcing its pivot to crypto in 2018 the company has raised nearly £5mn with share placings. It also placed shares in 2017, 2016, 2015, 2014, 2012, 2010 and 2009. More than 70 per cent of Quantum Blockchain stock sits in retail broker nominee accounts.
One conclusion to take from the AsicBoost saga is that, if any party finds a better way to mine crypto, the most profitable strategy is to mine lots of crypto. Publicising the method will only cause a fight about what the community considers fair and might prompt a redesign of the protocol.
That could help explain why an Aim-quoted company with a £14mn market cap has chosen to tease innovations that — if they work! — would upend the $1.2tn market for bitcoin. It doesn't help explain why a company with a money-printing machine would be selling stock rather than printing money. We’ve tried to understand the reasons. But apparently, we’re not allowed to know.