AI, technology & valuations

By Hyperion Asset Management CIO, Mark Arnold & Deputy CIO, Jason Orthman

Artificial intelligence (AI) represents not merely another innovation cycle, but a fundamental reimagining of how value is created, work is performed, and society functions.

Hyperion's Global Growth Strategy is uniquely positioned to capture the outsized returns we believe this transformation will generate over the coming decade. In addition, we believe Hyperion's Australian Growth Strategy maintains strategic exposure to high-quality software and platform businesses that are not only defensible against AI disruption but are actively leveraging AI to enhance competitive advantages, accelerate growth, and expand margins.

History provides clear precedent for how transformational technologies create wealth. Over the past few decades, we have witnessed four major innovation cycles: the internet in the 2000s, smartphones in the 2010s, cloud computing through the 2010s and early 2020s, and now AI combined with robotics. Hyperion has successfully captured alpha in each previous cycle by identifying and investing in the highest-quality structural growth companies.

In our view, AI will be the largest of these four cycles. We expect it will create more value and wealth for society than its predecessors, driven by its unique capability to not merely facilitate human activity, but to augment and, in many cases, replace human cognitive labour. Major medical and scientific breakthroughs, unprecedented productivity gains, autonomous transportation, and humanoid robots that eliminate dangerous and repetitive work will fundamentally reshape our economy.

The distinction between AI and previous innovations is critical. While the internet facilitated better communications, enabled e-commerce, and provided easy access to information, it remained fundamentally passive - a tool requiring human direction and judgment. AI possesses capabilities the
internet never could: thinking, planning, predicting, analysing, and solving complex problems autonomously. Digital intelligence, particularly when combined with physical embodiment through robotics, will prove substantially more valuable than the internet ever was.

We believe that comparisons to the dot-com bubble of 1999 fundamentally misunderstand both that era and the current moment. The average price-to-earnings ratio for top technology stocks in 1999 was approximately 67x earnings. Today, leading US technology companies trade at an average of 28x, less than half the valuation extreme of the dot-com bubble era.

More importantly, the qualitative difference is stark. Most leading stocks during the tech bubble were lossmaking concept companies with no revenue, no competitive advantages, and no credible path to profitability. They represented speculation on what the internet might enable, not businesses already
generating substantial cash flows from proven business models.

Today's technology leaders operate profitable enterprises with formidable competitive advantages, using digital intelligence to enhance existing business models while creating entirely new revenue streams. The return on investment from deploying AI is demonstrably attractive: businesses are achieving significant cost savings and productivity improvements, cloud providers remain highly profitable with accelerating growth, and some software companies are successfully monetising AI through premium subscriptions with expanding use cases.

Concerns about circular financing also appear overblown. For example, NVIDIA's ecosystem investments remain relatively immaterial compared to expected revenues from their next generation Blackwell and Rubin chips of approximately USD$500 billion. These investments are strategic, building a broader ecosystem for sustainable long-term growth rather than propping up artificial demand.

The hyperscalers - Microsoft, Amazon, and Google - have funded their entire AI infrastructure spend through operating cash flows while maintaining very high returns on capital. If AI economics were deteriorating, we would witness capital expenditure pullbacks. Instead, investment is accelerating, driven by strong customer demand and clear monetisation pathways.

The rate of AI adoption is historically unprecedented. ChatGPT achieved 40% US market penetration in just three years. By comparison, the internet and personal computers required ten years to reach similar penetration levels. Full adoption of large language models is forecast within ten years, compared to 25 years for the internet and PCs to achieve universal usage.

ChatGPT accumulated one million users within five days of its launch. Today, it serves 800 million weekly users, a scale that continues to expand exponentially. As large language models incorporate more sophisticated AI-derived tools, monetisation rates will accelerate rapidly. When we consider that Google has nearly 5 billion monthly active users and Facebook exceeds 3 billion, the revenue and earnings potential as AI platforms scale to comparable levels becomes clear.

This adoption velocity supports our thesis that digital agents will become prevalent across workplaces within years, not decades, with physical real-world AI applications following quickly behind.

Hyperion's Global Growth Strategy has been deliberately constructed to capture value creation across the entire AI technology stack, from foundational semiconductor manufacturing through cloud infrastructure to application-layer software and physical AI embodiments.

Foundation Layer: ASML Holding NV (ASML) provides the only extreme ultraviolet (EUV) lithography equipment capable of manufacturing advanced semiconductors essential for AI compute. As chips advance and AI capabilities expand, ASML's equipment becomes increasingly critical within fabrication plants globally.

Chip Architecture: NVIDIA Corp dominates GPU design for AI training and inference, while ARM Holdings PLC (ARM) provides market-leading CPU architecture at global scale. Initially, accelerated AI processing occurs within data centres; however, as use cases expand into physical AI - such as autonomous vehicles, humanoid robots, and edge devices - the number of GPUs required to process tokens, and the resulting power consumption, will multiply exponentially. ARM will specifically benefit as AI moves to edge devices where CPUs deliver superior cost and energy efficiency.

Cloud Infrastructure: The hyperscalers are experiencing remarkable growth acceleration. Amazon Web Services increased revenue growth from 12% two years ago to 20% currently. Microsoft's Azure cloud business achieved 40% revenue growth in the most recent quarter. This momentum will strengthen as inference compute - the actual revenue-generating deployment of AI models - scales to multiples of training compute requirements.

Application Software: Companies like ServiceNow, Inc. and Palantir Technologies Inc. (Palantir) are already monetising AI through tiered subscription models and AI-enabled products. Palantir exemplifies the extraordinary growth available to companies successfully embedding AI into enterprise workflows. When Hyperion initially invested in early 2022, Palantir aspired to 30% revenue growth. In the most recent quarter, Palantir reported 63% revenue growth, and US commercial revenue growth accelerated from 71% in Q1 2025 to 93% in Q2 2025 to 121% in Q3 2025.

Following exclusive access to Palantir's product demonstrations and discussions with their head of architecture in San Francisco, we gained crucial insights into the company's sustainable competitive advantages. Their proprietary ontology platform - which maps customer workflows and integrates siloed data and software systems - combined with forward-deployed engineers embedded within customer organisations, creates formidable barriers to competition. The integration of large language models into these workflows enables dramatically superior search and decision-making capabilities. We have significantly upgraded our terminal value expectations as Palantir demonstrates clear potential to become the enterprise software platform of choice in an AI-first world.

Physical AI: At the apex of the value stack sits real-world AI embodied in autonomous vehicles, robotic systems, and intelligent devices. Tesla Inc. (Tesla) leads this category with evidence mounting that the company has all but solved generalised autonomous driving. The company now operates robotaxi networks in Austin and San Francisco, with plans for US-wide expansion over the next 12 months. Tesla's competitive advantages - including vertical integration of hardware and software, avoidance of expensive LiDAR and high-definition mapping, and a global fleet of over eight million vehicles collecting 1.4 billion miles of proprietary data weekly - position it well to dominate autonomous transportation with superior unit economics.

The same AI architecture and hardware developed for autonomous vehicles will power Tesla's Optimus humanoid robot. These are entering production next year with an initial capacity of one million units annually, followed soon after by a second production line targeting ten million units annually. The implications for human labour markets are profound.

AI is adding measurable value across diverse applications: customer service automation, marketing optimisation, cybersecurity, content generation, code development, recommender engines, and advanced analytics. Meta Platforms, Inc. provides compelling evidence: AI-enhanced recommender engines have been the main driver of the company's accelerated revenue growth from 4% two-and-a-half years ago to 26% currently through more engaging, better-monetised content on Instagram and Facebook.

Amazon.com, Inc. demonstrates AI's impact on e-commerce through improved listing quality and assisted shopping features, leveraging its unique scale and first-party data to thrive in an AI-disrupted retail environment. These are not theoretical benefits, they are driving reported financial results today.

Axon Enterprise Inc exemplifies AI's potential in specialised applications. The company's Draft One product automates police report writing using audio and visual data from body-worn cameras. Given that police officers can spend approximately 40% of their time on report writing, the productivity unlock is substantial. This represents AI directly addressing a large, quantifiable inefficiency in a large industry.

We estimate that over 80% of Hyperion's Global Growth Strategy has direct optionality to emerging AI revenue streams. While most companies can benefit from AI through cost-line efficiency gains and improved productivity, only a select few possess the inherent capabilities and strategic intent to benefit from emerging AI revenue opportunities.

Even holdings not traditionally associated with technology benefit from AI. Companies across the portfolio are deploying AI for internal productivity gains, customer experience enhancement, and operational optimisation; benefits that flow directly to earnings without requiring new revenue streams.

We believe Hyperion's Australian Growth Strategy maintains strategic exposure to high-quality software and platform businesses that are not only defensible against AI disruption but are actively leveraging artificial intelligence to enhance competitive advantages, accelerate growth, and expand margins. Contrary to market concerns about AI posing an existential threat to incumbent software businesses, our analysis demonstrates that portfolio holdings possess three critical defensive attributes: demonstrated intent to embrace and innovate with AI, sufficient depth in workflows to create switching costs, and proprietary first-party data at scale that large language models cannot replicate through web searches alone.

Wisetech Global Ltd. (Wisetech) recently announced a transformational shift from its seat-based licensing model to a transaction-based commercial structure. Under the new CargoWise Value Packs model, customers will be charged per transaction rather than per user seat, fundamentally altering the business economics while simultaneously positioning Wisetech to capture the full value of AI-enabled productivity gains. These AI agents have the potential to enable Wisetech's customers to significantly reduce operating costs while increasing throughput. Wisetech believes that 50% of its customers' staff numbers could be eliminated within 2 years through the adoption of digital agents in freight forwarding and customs workflows.

Xero Ltd. has demonstrated clear intent to embrace AI through rapid product releases incorporating machine learning and automation. Technology One Ltd. has rapidly deployed AI capabilities across its flagship enterprise resource planning (ERP) and student management systems. REA Group Ltd. is aggressively investing in AI and has just begun its roll out of AI agents that can be employed by homeowners.

Following the technical correction in structural growth stocks through late 2025, valuation metrics have become increasingly compelling. The Global Strategy's forecast internal rate of return over the next decade sits at approximately 23% per annum pre-fees, which is above the long-term historical average.

The Global Strategy's projected earnings per share growth over the next decade stands at 25% per annum, elevated from historical levels. We believe the market is underappreciating the long-term growth and earnings profiles of our companies.

The Australian Growth Strategy's forecast 10-year internal rate of return sits at approximately 20% per annum pre-fees which is also above the long-term historical average, with the forecast EPS growth at 19% per annum over the next 10 years.

This widening gap reflects the market's systematic underestimation of the long-term earnings power of quality structural growth companies. It represents the type of opportunity that has historically generated the most significant alpha: a temporary disconnect between short-term market sentiment and fundamental long-term value creation.

While AI represents our highest-conviction theme, both Strategies (Australian and Global) maintain diversification across at least 9 distinct structural themes to manage risk appropriately. These include the shift from traditional retail to e-commerce, the transition from traditional media to digital platforms, the transition toward a sustainable economy, digital transformation of the workplace, modernisation of payments and banking, healthcare innovation, and others.

Furthermore, our Strategies are more diversified than surface-level analysis suggests because most portfolio companies operate multiple distinct businesses. The Domestic and Global Strategies maintain exposure to over 50 underlying business operations across many different industries, providing meaningful diversification while maintaining concentrated exposure to the highest-quality structural growth opportunities.

This approach reflects our understanding of power law dynamics in equity investing. Most listed stocks add no sustained value over long time periods. Historically, the majority of market returns have been generated by a few exceptional structural growth companies with the highest levels of innovation and structural growth drivers. We focus exclusively on identifying and holding these rare businesses.

We believe the evidence is clear and compelling: AI represents a paradigm shift, not a speculative bubble. Adoption is occurring faster than any previous technology in history. Monetisation is already evident across the value stack, from semiconductors through cloud infrastructure to application software and physical embodiments. Returns on investment are attractive and improving. The world's leading technology companies are profitable, generating substantial cash flows, and trading at reasonable valuations relative to both their own history and the opportunities ahead.

For investors willing to maintain a long-term perspective and tolerate short-term volatility, the current pullback represents an attractive entry point into what we expect will be the defining investment theme of the next decade. The hard work of building these exceptional businesses is well underway. The monetisation of AI is not theoretical - it is occurring today and accelerating.

We believe with continued double digit EPS growth and the release of AI agents in 2026, capital will return to leading SaaS companies that we own. Earnings growth drives share prices long term. With increased evidence of an AI strategy, P/E ratios should stabilise and potentially expand. The market has indiscriminately oversold some software businesses, due to disruption risk and we believe there has been a general capitulation in domestic quality structural growth names. This typically creates a compelling entry point to allocate additional capital. We witnessed a capitulation in large cap global quality growth in December 2022 (due to rising interest rates) which resulted in larger than normal returns in 2023 and 2024.