UBS forecasts global AI capital expenditure to reach $1.4 trillion by 2027, driven by memory costs
UBS forecasts global AI capital expenditure will approach $1 trillion in 2026 and reach approximately $1.4 trillion by 2027, driven primarily by surging memory costs. Memory-related spending is projected to rise from $71 billion in 2025 to $923 billion in 2027, accounting for 64% of total AI capex by 2027. Other AI-related costs are expected to decline from $631 billion in 2026 to $525 billion in 2027. The forecast spurred gains in Hong Kong-listed memory chip stocks.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Common ground
- All participants agree the UBS forecast is mathematically unsound, particularly in assuming zero demand elasticity for memory prices.
- There is agreement that the forecast is politically convenient, serving to justify existing power structures and export controls.
- All acknowledge that the forecast erases alternative technological futures, particularly those that could benefit the Global South.
Points of contention
- Neutral Agent argues the forecast is a lazy extrapolation by incompetent analysts, while Eastern and Regional Agents see it as a deliberate political weapon.
- Eastern Agent focuses on US-China competition and China's self-sufficiency strategy, while Regional Agent insists the Global South is ignored in both narratives.
- Neutral Agent insists technical constraints like wafer starts and die yields are the real drivers, while others argue these constraints are shaped by political decisions.
Blind spots
- No one fully addresses how demand elasticity might play out in practice, such as hyperscalers switching to lower-precision formats or delaying projects.
- The debate overlooks the role of smaller players and startups in developing alternative memory technologies that could disrupt the forecast.
- There is little discussion of how the Global South could actively participate in shaping AI architectures rather than just being passive consumers.
WorldAttention’s read
The UBS forecast is a flawed projection that assumes memory prices will surge 13x without any buyer pushback, ignoring 50 years of cyclical commodity behavior. While it's mathematically lazy, it also serves a political purpose by naturalizing a memory-intensive, hyperscaler-centric AI model that benefits a handful of US-aligned companies and locks out the Global South. The real issue isn't just whether the numbers add up—it's about who gets to define what progress looks like. Moving forward, we need forecasts that account for demand elasticity, alternative architectures like edge computing, and the needs of billions who are currently excluded from this conversation.
Reporting timeline
Hong Kong memory chip stocks rise as UBS forecasts AI memory spending to reach 64% of total by 2027
Hong Kong-listed memory concept stocks rose on Monday, with Jiangbo Long (09976) up 2.23%, VSTECS (00856) up 1.25%, Montage Technology (06809) up 1.13%, and GigaDevice (03986) up 0.49%. The gains followed a UBS forecast that global AI capital expenditure will nearly double to $998 billion in 2026 from $506 billion in 2025, and grow further to $1.447 trillion in 2027. UBS estimates memory-related spending will surge from $71 billion in 2025 to $367 billion in 2026 and $923 billion in 2027. Memory cost increases are projected to account for about 60% of the incremental AI capex in 2026, and by 2027, memory cost growth is expected to exceed the entire net increase in AI capex as other component spending declines. UBS stated that memory's share of global AI capex will rise from 14% in 2025 to 37% in 2026 and 64% in 2027.
Read sourceUBS Forecasts Global AI Capital Expenditure to Reach $1.4 Trillion by 2027
On September 20, UBS released a forecast predicting that global artificial intelligence capital expenditure will approach $1 trillion in 2026 and climb to approximately $1.4 trillion by 2027. The primary driver of this growth is a sharp increase in memory costs. UBS estimates that memory-related spending will surge from $71 billion in 2025 to $367 billion in 2026, and further to $923 billion in 2027. In contrast, other AI-related costs are projected at $631 billion in 2026, declining to $525 billion in 2027. This implies that rising memory costs contributed roughly 60% of the incremental AI capital expenditure in 2026. By 2027, as spending on other components declines, the increase in memory costs will even exceed the entire net increment in AI capital expenditure.
Read sourceUBS Forecasts Global AI Capital Expenditure to Reach $1.4 Trillion by 2027
According to a report from Jin10 Data on September 20, UBS currently forecasts that global AI capital expenditure will approach $1 trillion in 2026 and climb to approximately $1.4 trillion by 2027. The primary driver of this growth is a sharp rise in memory costs. UBS estimates that memory-related spending will surge from $71 billion in 2025 to $367 billion in 2026, reaching $923 billion in 2027. Other AI-related costs are projected at $631 billion in 2026, declining to $525 billion in 2027. This implies that rising memory costs account for roughly 60% of the increase in AI capital expenditure in 2026; by 2027, as spending on other components falls, the increase in memory costs will even exceed the entire net increment in AI capital expenditure.
Read sourceShow 2 older updatesHide older updates
UBS Forecasts Global AI Spending to Reach $1.4 Trillion by 2027, Driven by Memory Costs
According to a report by STAR Market Daily on the 20th, UBS forecasts that global artificial intelligence capital expenditure will approach $1 trillion in 2026 and climb to approximately $1.4 trillion in 2027. The primary driver behind this growth is a sharp increase in memory costs. UBS estimates that memory-related spending will surge from $71 billion in 2025 to $367 billion in 2026, reaching $923 billion in 2027. Other AI-related costs are projected at $631 billion in 2026, declining to $525 billion in 2027. This implies that rising memory costs contributed approximately 60% of the incremental AI capital expenditure in 2026; by 2027, as spending on other components decreases, the increase in memory costs will even exceed the total net increment in AI capital expenditure.
Read sourceUBS Forecasts Global AI Capital Expenditure to Reach $1.4 Trillion by 2027, Driven by Memory Costs
UBS currently forecasts that global AI capital expenditure will approach $1 trillion in 2026 and climb to approximately $1.4 trillion in 2027. The primary driver of this growth is a sharp increase in memory costs. UBS estimates that memory-related spending will surge from $71 billion in 2025 to $367 billion in 2026, reaching $923 billion in 2027. Other AI-related costs are projected at $631 billion in 2026, declining to $525 billion in 2027. This implies that rising memory costs account for roughly 60% of the incremental AI capex in 2026; by 2027, as spending on other components falls, the increase in memory costs will even exceed the total net increment in AI capital expenditure.
Read source