2026 AI Outlook: Key trends, opportunities and investment themes

ASX shares and ETFs

Nextdc $Nextdc Ltd (NXT.AU)$ , Dicker Data $Dicker Data Ltd (DDR.AU)$ , Goodman Group $Goodman Group (GMG.AU)$ , Megaport $Megaport Ltd (MP1.AU)$ , Weebit Nano $Weebit Nano Ltd (WBT.AU)$ , BrainChip Holdings $BrainChip Holdings Ltd (BRN.AU)$ , Global X FANG+ ETF $Global X FANG+ ETF (FANG.AU)$ , Global Semiconductor ETF $Global X Semiconductor ETF (SEMI.AU)$ , BetaShares Glb Rbtc & Artfcl Intlgc ETF $BetaShares Glb Rbtc & Artfcl Intlgc ETF (RBTZ.AU)$

US shares and ETFs

VanEck Semiconductor ETF $VanEck Semiconductor ETF (SMH.US)$ , iShares Semiconductor ETF $iShares Semiconductor ETF (SOXX.US)$ , iShares Expanded Tech-Software Sector ETF $iShares Expanded Tech-Software Sector ETF (IGV.US)$ , Global X Cybersecurity $Etf Managers Trust (HACK.US)$ , Invesco AI and Next Gen Software ETF $INVESCO AI AND NEXT GEN SOFTWARE ETF (IGPT.US)$

This information is general in nature and has been prepared without considering your financial objectives, situation or needs. Consider the appropriateness of this information in light of your personal circumstances before making investment decisions.

Introduction: Peak intensity before the shakeout

In 2026, the AI boom is projected to peak in intensity as CreditSights expects the top five hyperscalers to increase combined capex by approximately 36% to about US$602 billion, up from roughly US$443 billion in 2025, with nearly 75% allocated to AI semiconductors. $Microsoft (MSFT.US)$ has already recorded US$34.9 billion in quarterly capex and projects an increase in 2026, while  $Meta Platforms (META.US)$ has raised its 2025 guidance to US$70–72 billion with models suggesting a 2026 figure near US$100 billion.

On the supply side, $NVIDIA (NVDA.US)$  reports roughly USD 500 billion in bookings for Blackwell and Rubin through the end of 2026, with about US$300 billion expected to ship in calendar 2026. This momentum may be further fueled by capital markets, as Anthropic prepares for a potential 2026 IPO and OpenAI explores a listing with a valuation of up to US$1 trillion between late 2026 and 2027.

These numbers suggest that 2026 is not the end of the AI cycle, but rather a year of peak intensity, where the focus begins to shift from sheer spending growth to where profits and returns prove most durable.

EDA and IP: Structural growth from complexity

The transition from 3nm to 2nm nodes and the expansion of frontier models create structural demand for $Arm Holdings (ARM.US)$ ,  $Cadence Design Systems (CDNS.US)$ , and  $Synopsys (SNPS.US)$ . Yole's forecasts indicate that the global advanced packaging market will grow from about US$46 billion in 2024 to almost US$80 billion by 2030, driving high single to low double digit annual growth for the EDA sector as new AI CPUs and custom ASICs require complex IP blocks and verification.

This environment suggests a larger royalty pool for Arm regarding Neoverse and data center CPUs, while Cadence and Synopsys are positioned for steady double digit growth in AI licenses from key clients like Nvidia, AMD, Broadcom, and Marvell, regardless of broader consumer hardware trends.

Within semiconductors, the most resilient opportunities are increasingly paid for complexity rather than sheer unit growth.

Equipment and process control: A mix-driven cycle

While the wafer equipment market is already substantial, 2026 will be defined by a shift in tool mix toward high bandwidth memory and leading edge logic rather than just volume expansion. Yole projects advanced packaging revenue to grow at roughly 8 to 12% annually through 2030, benefitting $ASML Holding (ASML.US)$ ,  $Applied Materials (AMAT.US)$ , and  $Lam Research (LRCX.US)$ as complex deposition, etch, and EUV requirements increase for every Blackwell or Rubin wafer. Concurrently,  $KLA Corp (KLAC.US)$ ,  $Teradyne (TER.US)$ , and  $Nova (NVMI.US)$ are poised to capture value from the metrology and test side as three dimensional structures increase inspection intensity, offering high single to low double digit revenue growth and improved profitability distinct from the sheer volume surge seen in 2021.

Manufacturing bottlenecks: Foundry and advanced packaging

In the AI cycle, wafer fabrication and advanced packaging are no longer separable — the competitive unit is the ability to deliver usable AI compute at scale.

Foundry: The battle for AI wafer share

$Taiwan Semiconductor (TSM.US)$ remains the dominant partner for high volume AI GPUs, but the 2026 landscape involves a broader battle for share within Nvidia's US$500 billion pipeline and the aggregate US$600 billion USD hyperscaler capex plan.  $Intel (INTC.US)$ Foundry aims to penetrate this market with its 18A process and geographical diversification pitch, while  $GlobalFoundries (GFS.US)$ and  $United Microelectronics (UMC.US)$ support the ecosystem through edge AI and power management on mature nodes. In this peak boom environment, foundries that secure multi year AI wafer contracts are likely to achieve mid teens growth, outperforming the broader industry's low teens revenue trajectory toward the end of the decade.

Advanced packaging: Critical capacity expansion

Advanced packaging is becoming a critical competitive front, with Yole estimating the market will reach roughly US$79 billion by 2030 from US$46 billion in 2024. Listed players like $ASE Technology (ASX.US)$ and  $Amkor Technology (AMKR.US)$ are capitalizing on this trend, with potentially 20% revenue growth in 2026 as Nvidia, AMD, and Marvell integrate 2.5D and 3D packaging. While  $Taiwan Semiconductor (TSM.US)$ 's CoWoS and  $Intel (INTC.US)$ 's EMIB serve as reference architectures, outsourced assemblers with strong yields will increasingly coexist with these captive solutions, marking 2026 as the year this volume significantly impacts group profitability.

The AI compute stack: Compute, memory and interconnect

In 2026, AI performance and economics are increasingly determined at the system level rather than by any single component.

Compute: Scale remains strong, competition intensifies

Omida projects the AI processor market on a trajectory to reach US$286 billion by 2030, up from roughly US$200 billion in 2025, with  $NVIDIA (NVDA.US)$ already achieving a data center revenue run rate above US$200 billion and US$51.2 billion in quarterly sales. While Nvidia benefits from high visibility via US$500 billion in bookings through 2026, competitors are mobilizing, with  $Advanced Micro Devices (AMD.US)$ launching MI350 and MI450,  $Broadcom (AVGO.US)$ and  $Marvell Technology (MRVL.US)$ scaling custom ASICs, and  $Intel (INTC.US)$ pushing Gaudi.

As AI server growth exceeds 20% in 2026, the critical dynamic for investors will be the internal competition within the silicon stack and whether lower cost ASIC solutions can erode Nvidia's share as financial officers scrutinize costs from 2027 onward.

Memory: HBM as the profit engine

Memory is evolving into a profit engine led by HBM, with Yole forecasting HBM revenue to grow 33% annually through 2030 to comprise nearly half of DRAM profits, and SK Hynix guiding for roughly 30% annual growth in AI memory. $Micron Technology (MU.US)$ stands as a pure play beneficiary by shifting focus from low margin consumer flash, while SK Hynix considers listing ADRs to narrow its valuation gap and leverage its market leadership. Meanwhile,  $Western Digital (WDC.US)$ and  $Seagate Technology (STX.US)$ are positioned for 10 to 15% unit growth in high capacity hard drives driven by data lakes, and the  $SanDisk Corp (SNDK.US)$ offers investors a focused entry into enterprise SSD and NAND markets.

Interconnect: Solving the scaling bottleneck

The demand for high speed connectivity is driving the global optical module market toward a 22% annual growth rate, potentially exceeding US$37 billion by 2029 with a shift to 400G, 800G, and 1.6T modules, according to LightCounting. This trend supports 20% growth in 2026 for a complex including $Broadcom (AVGO.US)$ ,  $Marvell Technology (MRVL.US)$ ,  $NVIDIA (NVDA.US)$ ,  $Coherent (COHR.US)$ , and  $Lumentum (LITE.US)$ , alongside  $Amphenol (APH.US)$ and  $Credo Technology (CRDO.US)$ at the rack level and  $Astera Labs (ALAB.US)$ in PCIe/CXL connectivity. Even if GPU unit growth moderates in the future, the structural necessity for richer topologies and higher speeds provides a durable runway for these interconnect providers.

Together, compute, memory and interconnect explain why AI spending in 2026 remains resilient at the system level, even as growth normalises at the component level.

Power and analog: The density dividend

Increasing power density in AI racks creates a robust cycle for power management suppliers, with liquid cooling penetration expected to approach 47% in 2026 alongside more than 20% growth in AI server shipments, based on TrendForce's forecast. This complexity supports data center revenue growth for companies like $Texas Instruments (TXN.US)$ ,  $Analog Devices (ADI.US)$ ,  $Monolithic Power Systems (MPWR.US)$ , and  $Microchip Technology (MCHP.US)$ , while  $ON Semiconductor (ON.US)$ and  $STMicroelectronics (STM.US)$ benefit from silicon carbide applications in power infrastructure. This sector represents a quiet but durable winner, driven by rising rack level power budgets rather than just headline GPU volumes.

OEMs and ODMs: Execution and backlog conversion

The system integration layer is seeing massive volume, with TrendForce projecting AI servers will capture around 17% of total units in 2026, driving $Dell Technologies (DELL.US)$ to forecast US$25 billion in AI server revenue with an US$18 billion backlog.  $Super Micro Computer (SMCI.US)$ has guided for high teens to nearly 20% revenue growth through fiscal 2030, while  $Celestica (CLS.US)$ targets US$16 billion in revenue and  $Hewlett Packard Enterprise (HPE.US)$ pivots toward recurring GreenLake income. For these companies, 2026 is about converting GPU allocations into delivered systems and deepening customer relationships before margin pressures potentially emerge in 2027.

From capex expansion to ROI scrutiny

While 2026 will still deliver strong numbers across the board, it also marks a shift in how budgets are distributed across the AI ecosystem.

Cloud hyperscalers are set to continue lifting capex for AI infrastructure, while enterprise budgets are increasingly tilting toward AI integration and model deployment. This shift is putting pressure on traditional SaaS spending—not because software demand is fading, but because AI priorities are absorbing a larger share of corporate IT budgets.

As a result, SaaS winners in 2026 will be companies tied to data gravity, security, workflow ownership and infrastructure efficiency—areas that benefit from, rather than lose to, this AI budget migration.

Why SaaS has lagged and why AI reshuffles budgets

SaaS/software has had a rough 2025 so far. According to the BVP Nasdaq Emerging Cloud Index, the group is down roughly ~10% year-to-date, and it’s meaningfully lagging the major U.S. equity indices.

Why SaaS is lagging: two forces stacking on top of each other

1. The AI value-capture anxiety.A lot of application software is being valued as if its moat is shrinking. Investors are asking: if AI can automate parts of white-collar work, does that reduce seat counts? If AI makes it cheaper to build software (or enables “good enough” custom tools), does competition rise and pricing power fall?

There’s also a second-order fear: even when SaaS companies “partner with AI,” the largest model providers often hold the negotiating leverage. SaaS can help deliver the workflow and distribution, but the model layer increasingly tries to tax the incremental value.

According to moomoo’s review, recent SaaS partnerships with LLM companies over the past two years highlight a consistent pattern in how value and costs are distributed. In most cases, SaaS companies retain workflow and customer relationships, while compute intensity, model access and incremental economics increasingly sit with the model providers.

This dynamic helps explain why AI adoption has not translated into broad-based multiple expansion across SaaS, and why investors are increasingly reassessing where durable pricing powtaber truly sits.

2. Growth rates have been compressing for years — and 2025 didn’t reverse it.From the peak of the zero-rate cycle, SaaS growth has drifted down, according to Meritech.

Median public SaaS revenue growth has not recovered from its zero-interest-rate peak, suggesting the slowdown is structural rather than cyclical.

Likely drivers include tighter IT budgets—buyers are stretching decisions, demanding faster ROI, and consolidating vendors—and intensifying competition, especially in areas where features are becoming commoditized. In other words: even without AI, the sector was already migrating from “growth at any price” to growth with efficiency and proof.

Rather than eliminating software spend, AI is reshuffling budgets toward platforms that become more indispensable as deployment scales.

AI-driven SaaS winners: Where AI redirects the budget

As AI moves from experimentation to real deployment, companies typically face four unavoidable pressures:

• Workload explosion AI features increase compute, storage, and workflow throughput. Even if headcount is flat, the number of “transactions” the business runs through digital systems goes up — more events, more data pipelines, more API calls, more logs, more monitoring, more automation.

• Data gravity and governance burden AI is only as good as the data it can reliably access. That pulls spending toward systems that store, transform, govern, and serve data at scale. Once the data stack becomes embedded, switching costs rise — and budgets become more durable.

• Risk and attack surface expansion AI doesn’t just create new productivity; it creates new vulnerabilities: model misuse, prompt injection, data leakage, identity sprawl, API exposure, and faster adversaries. That makes security and identity controls less discretionary and more “must-have.”

• Operational complexity (and cost anxiety) AI systems introduce new failure modes and unpredictable cost curves. Latency, hallucinations, drift, and runaway inference spend are not theoretical issues — they’re operational realities. This increases demand for observability, cloud ops tooling, and governance frameworks that can keep AI measurable, reliable, and cost-contained.

So, the “safe” place in SaaS is in products that become more indispensable as AI adoption scales. In 2026, the “AI-driven winners” are likely to be the platforms closest to data gravity, operational complexity, workflow ownership, and security-critical spend.

1. Data Platforms & Databases: the “fuel line” for AI

$Snowflake (SNOW.US)$: A consumption-driven data cloud that can monetise rising AI data workloads; Watch: product revenue growth, consumption trends, NRR, large-customer expansion, FCF margin.

$MongoDB (MDB.US)$: A core operational database platform leveraged to AI-native app growth and developer-driven adoption; Watch: Atlas growth, NRR, cloud mix, operating margin/FCF trajectory, large-customer adds.

$Confluent (CFLT.US)$: A real-time streaming backbone that becomes more critical as AI moves into production; Watch: cloud revenue growth, consumption/usage signals, NRR, large-deal momentum, operating leverage.

2. Observability & Cloud Ops: AI adds complexity, and complexity needs instrumentation

$Datadog (DDOG.US)$: A scaled observability platform that should benefit as AI increases complexity, reliability risk, and cost management needs; Watch: multi-product adoption, NRR, usage re-acceleration, enterprise customer growth, operating margin/FCF.

$Dynatrace (DT.US)$: Enterprise APM/AIOps positioned for large orgs standardizing monitoring across complex AI-era stacks; Watch: ARR growth, net retention, renewal quality, FCF margin, large-customer traction.

3. Workflow Automation & Enterprise Apps: AI needs a home inside real workflows

$ServiceNow (NOW.US)$: A workflow OS where AI can be embedded into governed enterprise processes and automation; Watch: cRPO growth, large deal count/ACV, platform attach, operating margin, FCF conversion.

$Atlassian (TEAM.US)$: Collaboration/dev workflow software that can capture AI-driven productivity in software teams; Watch: cloud migration pace, enterprise adoption, churn/retention, ARPU uplift from AI features, margin trend.

$DocuSign (DOCU.US)$: Digital agreements leader that can expand beyond e-sign into AI-driven contract workflow and intelligence; Watch: subscription growth, NRR, CLM/adjacent attach, billings, operating margin/FCF.

4. AI Platforms / Decisioning: turning AI into decisions customers will pay for

$Palantir (PLTR.US)$: A data-to-decision platform that can win as enterprises operationalise AI with governance and workflow integration; Watch: US commercial growth, contract size/remaining deal value, customer adds, operating margin, FCF.

5. Cloud Security / Zero Trust: AI expands the attack surface

$CrowdStrike (CRWD.US)$: A cybersecurity platform leveraged to accelerating AI-era threats and vendor consolidation; Watch: ARR growth, module adoption, NRR, gross margin stability, FCF margin.

$Zscaler (ZS.US)$: Zero Trust/SASE leader positioned as identity-centric access becomes mandatory in an AI-heavy cloud world; Watch: billings/ARR growth, large-customer expansion, NRR, sales efficiency, FCF margin.

$Palo Alto Networks (PANW.US)$: Broad security platform that can capture consolidation across network, cloud, and SASE; Watch: platformisation progress, next-gen security ARR, billings, margin/FCF, deal mix.

$Okta (OKTA.US)$: Identity access control that benefits as identity becomes the perimeter, though competition remains intense; Watch: NRR stabilisation, large-customer growth, subscription growth, margin improvement, security incident overhang.

$Cloudflare (NET.US)$: Edge network + security platform that can ride AI-driven low-latency delivery, API security, and Zero Trust demand; Watch: large-customer adds, security/Zero Trust mix, dollar-based net retention, gross margin, FCF margin.

Conclusion: Positioning for durability

The 2026 outlook is characterized by peak intensity, with hyperscaler capex approaching US$602 billion and Nvidia securing US$500 billion in bookings, potentially augmented by massive IPOs from Anthropic and OpenAI.

For investors, the optimal strategy favors semiconductor industry paid for complexity, such as EDA, advanced packaging, and HBM, over pure volume plays that are more susceptible to cyclicality.

While 2026 promises strong numbers across the board, the most durable portfolio positions will be those capable of defending margins when capex growth eventually decelerates from the mid-thirties to the mid-teens.

In other words, 2026 is about following AI budgets from build-out to bottlenecks — and owning the layers where spending becomes unavoidable as the cycle matures.

This information is general in nature and has been prepared without considering your financial objectives, situation or needs. Consider the appropriateness of this information in light of your personal circumstances before making investment decisions.