Tariff Truce

by Ryan Shea

Digital asset markets returned to their winning ways last month as evidenced by the 9% rally in our flagship large cap Top10 Crypto CTI. The catalyst for the bullish price move was a substantial easing in global trade war fears instantiated by Trump’s Liberation Day tariff announcement back in early April.

Remarkably, despite all the pessimistic takes about the US tariffs, a benign macro outcome was always a very distinct possibility given Trump prides himself on being a strong and tough negotiator. Indeed, this is exactly why in the previous monthly update we wrote…

“… it is perfectly plausible to assume that by announcing large and widespread tariffs – at the risk of triggering Great Depression 2.0 – Trump deployed military style “shock and awe” tactics to bring other nation states to the negotiating table with the intention of getting them to remove the pre-existing trade impediments on US exporters in return for being exempt from US tariffs. Certainly, this is a more plausible explanation than the intellectually lazy option many took, which was to simply dismiss Team Trump as a bunch of economic imbeciles for announcing such draconian trade measures”.

As it transpired, betting against the lazy option turned out to be a profitable trade last month.

Trade War Averted

The first indication that a benign outcome was the more likely outcome came on May 8, when the UK became the first country to sign a bilateral trade deal with the US. However, what really invigorated the animal spirits of crypto investors was the news four days later that the US and China had committed to cutting reciprocal tariffs by 115 percentage points for a period of 90 days. (Trump adopted similar tactics versus the EU after he delayed the introduction of 50% tariffs for the region until July 9.)

While only a pause, and notwithstanding the fact that tariffs are set to remain above their pre-Liberation Day levels (10% for the US and 30% for China¹ – assuming the administration finds a way to circumvent the recent decision by a US trade court to block the Trump tariffs) the partial rollback sent a clear signal that neither country wanted to risk an economic “decoupling” – an outcome that would have constituted a potentially very negative macro tail event. Hence, the rebound in risk appetite, which provided the macro foundation for last month’s crypto rally. (NB: Alongside this cyclical tailwind, crypto also gains from the growing caution among other nation-states in relying on the US dollar as a reserve currency—a hesitation rooted in the erosion of trust caused by Trump’s brinksmanship negotiation tactics, which persists even though the immediate threat of a trade war has passed.)

Alt-Coin Season Incoming?

As we noted previously, one aspect of the current bull market that is rather unusual is Bitcoin has led from the front. This is evidenced by its share of overall crypto market cap – a metric known as Bitcoin dominance – having steadily risen from a low of around 40% at the end of the last bear market to a high of 64% in April – see chart.

Bitcoin Dominance vs. Price

Source: Trading View

As to why the crypto market is showing such atypical internals, the most likely explanation is the surge in institutional demand for the seminal cryptocurrency facilitated by the green lighting of spot Bitcoin ETF in the US in January 2024. In the 17 months since then, these products have collectively attracted capital inflows of more than $40bn (see image) – a feat that took Gold ETF six years to achieve.

Bitcoin Spot ETF Cumulative Flows – Total (USDm)

Source: Farside Investors

That said, maybe there is a subtle shift taking place because even though Bitcoin went on to post a new all-time high above $111,000 other cryptocurrencies outperformed the market leader during last month’s bullish move. As a result, Bitcoin’s dominance edged down slightly. Clearly, extrapolating from such a small shift is dangerous (squint and you can see the uptick in the following chart), but it could be tentative evidence that the “alt-coin season” is finally in the offing.

Token Dominance vs. BTC Price

Source: Glassnode

Eyes On Ethereum

One of the top performers last month was second placed Ethereum, a cryptocurrency that has been beset with issues over the past several quarters. The successful roll out of the Pectra upgrade on mainnet on May 7 unquestionably boosted sentiment towards the token. Included in the upgrade was EIP-7702 that allows externally owned accounts to adopt smart-contract behaviour temporarily, permitting the batching of transactions, the ability to pay gas fees in tokens other than ETH, and a fund recovery mechanism when a private key is lost.

The other significant change introduced by the Pectra upgrade was the raising of the validator stake limit from 32ETH introduced back in 2022 when Ethereum transitioned from proof-of-work to proof-of-stake (aka The Merge), to 2,048 ETH. Such a substantial increase makes the node management process much easier for larger stakers and hence makes the Ethereum blockchain more institution-friendly, a well-timed change given how increasingly crypto-curious many corporate entities are becoming.

Vitalik’s Vision

Just prior to the Pectra upgrade Ethereum co-founder Vitalik Buterin published a post on his blog outlining his latest thoughts as to how Ethereum can succeed in becoming “the world ledger: the platform that stores civilization’s assets and records, the base layer for finance, governance, high-value data authentication, and more”.

Achieving such an ambitious goal obviously requires scalability, the element in the Blockchain Trilemma that typically gets sacrificed², but as the post mentions it also requires an additional feature: resilience. According to Vitalik, an important aspect of this objective is the protocol’s simplicity. This is not a word one often hears in relation to Ethereum, but is very much of a defining feature of its predecessor Bitcoin, as Vitalik readily acknowledges in the post.

Right from the beginning, Ethereum had much greater ambitions than Bitcoin, but achieving this outcome generated a lot of additional protocol complexity – something that Vitalik admits was often due to his decisions – and this did not always payoff. Moreover, its structure means that it is unable to match the performance of newer blockchains (think Solana or NEAR) in terms of transaction bandwidth/scalability.

Almost certainly, the post was a response to criticism from many ETH bag holders that under the stewardship of Ethereum Foundation (EF) – the non-profit that helps guide protocol development – Ethereum was a blockchain losing its purpose and hence in need of revitalizing (re-vitaliking perhaps?).

With his latest proposal, Vitalik is following the tried and tested KISS (Keep It Simple) approach. To this end, he is proposing methods to simplify the consensus layer (how users agree on the state of the world), the execution layer (replacing EVM with another virtual machine or RISC-V – binary machine code that performs basic operations), shared protocol components (reuse code that does similar tasks) all while maintaining the desirable property of backwards compatibility. This is no small task, but it is probably the best way to ensure that Ethereum stays relevant in the ever-evolving crypto ecosystem.

Certainly, Vitalik’s latest vision for Ethereum met with a positive response from crypto investors and, together with the successfully roll out of the Pectra upgrade on mainnet, this helped Ethereum break out of its multi-year downtrend against Bitcoin on its way to posting an impressive 40% monthly gain.

AI HotSpot

While Ethereum performed well last month, the segment of the crypto that has attracted most attention of late is AI-related tokens. These tokens have significantly outpaced the broader crypto market over the past several weeks, although they did succumb to a bout of profit-taking in May, leaving the AI CTI up 5% and the AI Agents CTIs flat over the month. This relative strength reflects growing public recognition of AI’s transformative potential —fueled by the exponential improvement of the underlying models and a rapidly expanding set of real-world applications. After all, think about how much progress has occurred in the two and half years since the initial public release of ChatGPT, which is when AI really went mainstream.

Impressive as the OpenAI chatbot was at launch (based on GPT-3.5), the latest version (GPT 4o³) has improved such that it has an estimated IQ of over 130. This exceeds the threshold required to enter Mensa (the famous high IQ society) placing it in the top 2% of the population by intelligence. These conversational AI agents may not constitute artificial general intelligence (AGI) but from an investment perspective that is beside the point. What matters is this: for the cost of BTC 0.00000098 a month (or $20 in old money), anyone in the world with an internet connection has access to a digital “genius” assistant at their disposal 24/7. Moreover, the LLMs powering these AI agents are only getting better over time – see image below.

LLM IQ Test Results – Mensa Norway Quiz

Source: www.trackingai.org

Agency Incoming

One area where it is becoming increasingly clear that AI/AI agents will dominate is as the interface of human-digital interactions. “Googling” something will be out, and in its place everyone will be “GPT-ing”⁴. Not only that, but the use-case will go way beyond a simple information search. AI Agents will be given agency so as to be able to carry out instructions independent of direct human control eg, “Please book me a two night stay in a nice hotel in Rome near the Spanish steps this weekend. Include flights but nothing too early or late. Thanks⁵.”.

This trend stands to be crypto-positive because AI agents do not have a real world identity, making them ineligible to access tradfi banking services directly given the stringent KYC/AML criteria required by regulators. Of course, it is theoretically possible for their carbon-based owners to provide AI agents with access to their tradfi bank accounts to permit the financial transactions necessary for them to complete their tasks, but this is both risky and problematic. What if the AI agent made a financial transaction in error, or worse went rogue and started disbursing your funds all over the world? Also, payment card industry compliance not only places substantial restrictions on how and where card numbers are saved but storing security codes/pins is explicitly banned for rather obvious fraud reasons. This is the problem with the “human-first” design of tradfi payments systems.

An alternative, more financially secure option that circumvents the tradfi sector entirely, is for AI agents to utilize crypto payment rails. Smart contracts would be able to define the behaviour of AI agents and the funds required for them to make financial transactions on behalf of their carbon-based owners could be restricted to specific pre-funded digital wallets. Worse case, a rogue AI empties one of these digital wallets but, unless one recklessly chooses to put all of one’s crypto holdings in it, the losses would be limited.

Heaven Or Hell?

The above argument is a well-known AI narrative in the crypto industry but that is not the only way widespread adoption of AI technologies will impact digital asset prices. Arguably, the greatest impact will be driven by how AI affects the global economy. It is widely agreed that AI will have a dramatic impact yet – and somewhat amazingly given the speed of development – there is considerable disagreement as to whether this impact is a net positive or a net negative!!!

Advocates dream of a world of super abundance where the cost of goods and services collapses because AI augmented humans are able to provide the same level of services at a fraction of the cost. This productivity miracle would significantly boost potential GDP growth rates – the amount of economic growth that can be achieved without generating upward pressure on prices ie, causing inflation – greatly boosting living standards⁶. Freed from the bondage of labour, this would permit humankind to engage in more noble pursuits.

It is perhaps worth noting that such sentiments have been expressed before. Keynes in his 1930 essay “Economic Possibilities for Our Grandchildren” predicted that in a hundred years time (ie by 2030) people would be working no more than three hours a day : an observation that does not yet match the reality for the vast majority.

The reason for this mismatch is because historically as technology displaces some workers, it also leads to the creation of new jobs, which are able to absorb the excess supply of labour. In fact, according to a 2022 research paper by economist David Autor, more than 60% of jobs done in the US in 2018 had not yet been “invented in 1940”, meaning 85% of employment growth over the past 80 years came from jobs created by technological innovation.

Says Law Says…

Such a finding is in keeping with one of the lesser known laws of economics named after the French economist Jean-Baptiste Say, which in its most succinct form states that “supply creates its own demand”. This argument was nicely outlined in a 2023 blog post by Marc Andreessen entitled “Why AI Will Save The World”. To wit,

“When technology is applied to production, we get productivity growth – an increase in output generated by a reduction in inputs. The result is lower prices for goods and services. As prices for goods and services fall, we pay less for them, meaning that we now have extra spending power with which to buy other things. This increases demand in the economy, which drives the creation of new production – including new products and new industries – which then creates new jobs for the people who were replaced by machines in prior jobs. The result is a larger economy with higher material prosperity, more industries, more products, and more jobs.”

As Andreessen correctly observes this has indeed been the historical experience. The rolling out of new technologies has resulted in the cost of goods and service falling as the unit cost of production drops, freeing up income for consumers to purchase new goods and services, which serves to absorb workers displaced by the adoption of technology. That said, there are two key differences between AI and previous technologies.

First, AI has much wider applicability. The image below, taken from a recent IMF Staff Discussion Note, shows the degree to which employment shares are exposed to AI. It suggests that widespread adoption of AI could impact up to a third of workers in advanced economies (high exposure/low complementarity⁷), which if it resulted in their being fully displaced would constitute an economic shock equivalent in scale to the Great Depression when unemployment rates topped 25%!

Source: IMF

A Complimentary Robot

Moreover, this could be a low-ball estimate considering another area of rapid technological improvement – robotics (see image). According to a recent article in the UK Guardian newspaper…

“..[s]ome careers are obviously safe from robot takeover. Taylor Swift is not in danger. Nor is Harry Kane. Nor, for that matter, is Keir Starmer, or the as-yet-unnamed next archbishop of Canterbury. Famous artist, sportsperson, politician, priest – perhaps the four jobs that are the most resistant to automation. Unfortunately they’re not open to all of us.”

Robotics is extremely complementary to AI and combined these two technologies have the potential to disrupt not just one or two industries but a huge swathe of them, meaning both white collar and blue collar jobs are at risk. AI can, after all, already produce music or works of art that people deem meaningful (bye bye, Taylor Swift); robots display physical skills that top athletes would be hard-pressed to replicate (bye bye, Harry Kane); as for politicians the less said about those the better (bye bye, Keir Starmer).

Importantly, this change will occur at light speed when measured on the political/economic timescale. Consider the following scenario outlined by Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean by the AI Futures project that forecasts the future of AI. Based on their experiences working for OpenAI, the Centre for AI Policy, and as VC investors in AI companies, they predict that by October 2027 – as far into the future as the public release of Chat GPT was in the past – there will be 330,000 superhuman AI researcher agents each capable of thinking at 57X human speed – see image.

AI 2027 – A Prediction

Source: ai-2027.com

Given the lead author’s track record of predicting the pace of AI improvement since 2021 has been better than almost everyone else, this is a plausible scenario. It implies that in little over two years millions of workers could have already been displaced, and a billion or so more will by then realize that similar fates await them in the not-too-distant future. So, while it is tempting to dismiss those worried about the economic impact from the widespread adoption of AI/robotics as just modern-day Luddities – see image – such concerns are well-founded.

Luddite Reincarnated

Source: ChatGPT (obviously!)

Importantly, even if the aforementioned surge in productivity leads to a sharp fall in the price of goods and services that supports living standards for everyone as a whole, the fiat money system being debt-based, is not well designed to accommodate an economic shock where millions of people lose their jobs (and hence their ability to service their debts) in a relatively short space of time.

As AI and robotics fundamentally reshape the demand for human labour, societies will need to rethink not just debt and monetary systems, or income models, but the very definition of work itself in order to maintain stability and fairness. For sure, it will require much more than the establishment of universal basic income (UBI) schemes one often hears being discussed in this context. Embracing a form of money well-suited to a digital dominant age, whose value is not backed by debt would certainly be a step in the right direction.

Now, if only such a thing existed…

1 The same is also true of the UK even after it’s trade deal. In fact, Aside from the carve outs for British steel, aluminum and car exporters (the 25% tariff being cut to 10% up to a limit of 100,000 vehicles) in return for eliminating the UK tariff on US ethanol imports and a commitment to increase imports of US beef, UK exports are still subject to a 10% tariff. This suggests the Trump administration is still committed to action that will serve to reduce its external deficit.

2 The other two are security and decentralization. For more on the trilemma – see: https://trakx.io/resources/research/drivechain-the-latest-controversy-to-impact-the-bitcoin-community/

3 Higher tier OpenAI customers have access to ChatGPT 4.5, which has an estimated IQ range of 130-150. For reference Albert Einstein had an IQ of 160.

4 Due to OpenAI’s strong brand recognition, we expect this denominalization – or verbing – to be used even if many people choose AI agents hosted on decentralized open source AI platforms because they offer greater privacy, resilience – that word again! – and transparency.

5 Sam Altman has complained that including please and thank you in AI prompts costs OpenAI millions of extra dollars, but some AI researchers have found that being polite improves the AI outputs so it is worth doing – see: https://x.com/BrianRoemmele/status/1654180874894692367

6 A substantial boost to potential GDP due to AI boosting productivity, would make government debt metrics appear much more sustainable than is presently the case. To the extent that crypto holders have purchased tokens as a hedge against a fiscally incontinent collapse of the fiat money system, this could be deemed a crypto-negative: something to thinking about!

7 Complementarity is an estimate as to how much AI would be helpful to a human in performing their functions. The higher the degree of complementarity the more productivity enhancing it is as opposed to labour demand destructive. Of course, these benefits are, critically, contingent on employees possessing the necessary skills to utilize AI – see: https://www.imf.org/en/Publications/WP/Issues/2023/10/04/Labor-Market-Exposure-to-AI-Cross-country-Differences-and-Distributional-Implications-539656