Hyperion dismisses AI bubble fears, backs 25% earnings growth

As markets grapple with whether AI represents a speculative bubble or a genuine inflection point for long-term value creation, Hyperion Asset Management has planted its stake firmly in the latter camp and is betting big on the paradigm shift.

In a recent webinar, CIO Mark Arnold and Deputy CIO Jason Orthman argued that the current cycle bears little resemblance to the tech bubble that burst 25 years ago and long-term performance numbers make it hard to argue with the Brisbane-based fundies. Their Hyperion Global Growth Companies Fund - a portfolio embed with global AI and machine learning beneficiaries - has delivered +20% returns per annum since inception in 2014.

But with AI stocks commanding premium valuations, “cashless” companies and debt-fuelled spend, and sceptics drawing parallels to 1999, it begs the question - when will the economics of AI begin to pay off?

Bubble or next inflection point? Look at the numbers

The PE ratios tell part of the story, Arnold argues. Leading tech stocks traded at an average of 67 times earnings in 1999, compared to 28 times today. More importantly, today's companies have something their predecessors lacked: actual profits.

"Today's tech leaders have profitable businesses with strong competitive advantages that can use digital intelligence to enhance their existing business models and long-term growth profiles," Arnold said. "This was not the case in 1999."

The firm points to accelerating revenue growth across its holdings as evidence AI demand is real. Meta's revenue growth has jumped from 4% two and a half years ago to 26% currently. Amazon Web Services has moved from 12% to 20% growth over the same period.

Palantir (NASDAQ: PLTR) is a standout in its portfolio. When Hyperion first bought the stock, management targeted 30% revenue growth. The company is now delivering 63%, up from 18% two years ago.

The global companies leading AI monetisation

Hyperion's framework for AI monetisation spans five layers, from chip manufacturing through to physical robotics.

At the foundation sits ASML (NYSE: ASML), the sole provider of extreme ultraviolet (EUV) lithography equipment. Above that, Nvidia dominates GPUs, whilst ARM controls CPU architecture. The cloud providers - AWS, Azure - capture usage growth as AI inference scales. Software companies like ServiceNow and Palantir build AI-enabled applications on top.

The physical AI layer represents the longest-term opportunity. "As we move to physical AI, we expect the number of GPUs, tokens and watts to increase multiples over the next 10 to 20 years," Orthman said. 

"Ultimately, AI will move from the data centres onto our edge devices, into our cars, into humanoid robots, and this is where ARM will really benefit."

Tesla (NASDAQ: TSLA) features prominently in this vision as “the leader in real-world AI.” Arnold recently tested the company's full self-driving software in Brisbane and reported the system handled most driving tasks independently. 

Tesla plans to launch Robotaxi services across the US over the next 12 months, following Waymo's success in capturing nearly 30% market share in San Francisco within two years.

Adoption at unprecedented speed

ChatGPT reached 1 million users within five days of launch and now counts 800 million weekly users. It took 10 years for the internet and personal computers to reach 40% adoption. AI achieved the same penetration in three years.

Hyperion expects full adoption within a decade, compared to 25 years for earlier technologies.

"This supports our thesis that digital agents will be prevalent across the workplace in the coming years and physical real-world AI will quickly follow," Orthman said.

The circular economy question

Critics have questioned whether AI represents a circular economy, with companies investing in each other's infrastructure without genuine end-user demand.

Hyperion dismisses this concern. Nvidia's ecosystem investments totalled roughly $24 billion this year, small relative to the expected $320 billion in revenues from its Blackwell and Ruben chips, which carry approximately 50% margins.

The hyperscalers - Amazon, Microsoft, Google - have funded their entire AI infrastructure spend through operating cash flow whilst maintaining high returns on capital. Revenue growth rates have accelerated, not slowed.

Domestic market challenges

On the domestic front, market-darling growth stocks of the ASX over recent years have encountered headwinds in 2025. Over the past six months, quality companies have underperformed loss-making businesses as investors rotated out of technology and healthcare into resources and early-stage exploration.

Macquarie data presented by Orthman shows quality compounders generate significant long-term returns whilst loss-makers destroy value. Yet recent flows suggest investors are chasing short-term momentum over fundamental quality.

"From our perspective, that's not a sustainable way to create wealth," Orthman said.

The firm sees software concerns as overblown. Companies with genuine competitive advantages, such as proprietary data at scale, deep workflow integration, and proven ability to innovate, should withstand disruption from large language models.

3 ASX quality compounders

1. WiseTech (ASX: WTC) attracted selling pressure following founder Richard White's personal issues and broader AI fears. Hyperion views the selloff as overdone, pointing to the company's new transaction-based pricing model and exclusive first-party data for training AI agents.

2. Xero's (ASX: XRO) decline also appears excessive given the company's execution on its strategic plan and the strategic logic behind its Lilo acquisition for competing in the US market.

3. Life360 (ASX: 360) delivered another strong result, though the market focused on a single quarter's US monthly active user metric. Orthman argues this misses the bigger picture. Less than 10% of the user base currently pays for subscriptions, leaving significant room to monetise through advertising.

The opportunity cost

Hyperion's 10-year earnings forecasts highlight the gap between its outlook and market expectations. The firm projects 25% annual EPS growth for its global strategy, 19% for Australian growth, and 23% for small caps.

The "time arbitrage gap" - the difference between Hyperion's five-year EPS forecasts and consensus - sits at 30% for the global fund, 28% for Australian growth, and 16% for small caps.

Internal rates of return currently range from 20% to 23% across the three strategies, well above historical averages.

"The recent technical correction in structural growth stocks is providing an attractive opportunity for long-term investors," Arnold said.

Whether AI justifies these growth expectations remains the central question. Hyperion's case rests on profitable business models, accelerating adoption, and expanding use cases. The tech wreck comparison, they argue, misses the fundamental difference: this time, the technology works and companies know how to make money from it. 

Managed Fund

Hyperion Global Growth Companies Fund - Active ETF

Global Shares

View

Managed Fund

Hyperion Australian Growth Companies Fund

Australian Shares

View

Managed Fund

Hyperion Small Growth Companies Fund

Australian Shares

View