~With contributions from Vivien Lee, Aaron Ma and Eric Savoie
April webcast
Our monthly economic webcast for April can be found here: “Oil in the crosshairs.” Please note that it was recorded on April 1 – prior to the April 7 Iran-U.S. ceasefire announcement.
Iran war
The pendulum swings
The sentiment pendulum continues to swing. After a month of nearly relentless pessimism about the war with Iran, a two-week U.S.-Iran ceasefire was announced on April 7, greatly increasing optimism that lasting peace would soon be at hand, and sending stocks soaring and oil prices plummeting. The U.S. indicated that a 10-point plan proposed by Iran represented a reasonable framework upon which to negotiate.
But given that Iran’s framework includes recognizing Iran’s control of the Strait of Hormuz, accepting Iran’s right to nuclear enrichment and compensating Iran for war damages, striking an actual deal was always going to be considerably more difficult.
And so it is not a great surprise that negotiations over the April 11-12 weekend, with Pakistan as mediator, have not amounted to a finalized agreement.
Instead, the pendulum is now swinging back in the belligerent direction, with the U.S imposing its own blockade on the Strait of Hormuz. To the extent commerce through the Strait had already declined to barely a trickle due to similar Iranian efforts, the U.S. focus is on preventing Iran from exporting or importing its own products. (Iran had naturally been allowing its own vessels though its blockade.) Iran responded to the U.S. plan with the threat of attacking non-Iranian ports in the Persian Gulf. Oil prices have accordingly rebounded (see next chart).
Crude oil prices surge in the wake of the Iran war
As of April 13, 2026. Sources: Bloomberg, RBC GAM
This latest deterioration is of course unwelcome for the global economy. A rapid resolution to the war and the normalization of the price of oil is the desired outcome. The timing is unfortunate. The Strait of Hormuz had just started to unstick as Iran had allowed a variety of vessels connected to Pakistan, China, Iraq and even France to transit the area (see next chart).
Ship crossings in Strait of Hormuz collapsed, was reviving
As of April 12, 2026. Sources: Bloomberg, RBC GAM
Measured optimism persists
But there is still reason for measured optimism:
Both parties clearly want the war to end.
Both agreed to the temporary ceasefire.
Both are laying out conditions for what peace would look like.
They are actually meeting with one another. The ceasefire has also broadly held for the first week. The U.S. has not been attacking Iran and Iran has not been attacking the U.S., Israel or the Gulf states. The next chart shows an AI-based model we built that tracks Iran’s missile and drone usage.
Daily count of missiles and drones launched by Iran
As of April 13, 2026. Data compiled by RBC AI tool web scan of news reports, third-party analyses, and ministry of defense official statements from Gulf countries. Source: RBC GAM
Further good news is that China was apparently instrumental in bringing Iran to the table – so Iran’s political calculus depends on more than just whether the regime can survive a lengthier war. Late-breaking news reports re-emphasize China’s willingness to play a constructive role in resolving the war. If the U.S. succeeds in its blockade, China would be among the most affected countries given that it has continued to receive a significant amount of energy from Iran – further increasingly the likelihood that China will push for a speedy solution.
The very fact that the U.S. has escalated via an economic action rather than a ground invasion is also arguably a significant win.
It may take several rounds of negotiations before a deal is struck, but a deal should be possible. Using Iran’s 10-point peace plan as the framework, there is in our view room for a compromise deal (see next table).
Iran’s 10-point peace plan
Note: IAEA is the International Atomic Energy Agency. Sources: Bloomberg Economics, RBC GAM
As shown in the right-most column in green, relatively uncontroversial aspects of a deal would likely include:
The promise of nonaggression against Iran, and the end of the war on all fronts – including in Lebanon.
A reduction of U.S. combat forces in the region – mainly unwinding the recent build-up, without abandoning existing American bases in the region.
Lightening Iranian sanctions, both directly and via UN resolutions. But nuclear oversight via the International Atomic Energy Agency (IAEA) probably continues.
No direct compensation to Iran for war damages (though revenue via a Strait of Hormuz toll could be interpreted in this light).
Conversely, and depicted in yellow, trickier items include:
Does Iran get to continue controlling the Strait of Hormuz? We are inclined to think the answer will either be “no” or if the U.S. becomes desperate for an agreement, it might accede to the Strait of Hormuz being jointly managed with Oman or possibly with the U.S. Any formalized control of the flow of goods through the Strait would represent an unfortunate development. But at the proposed Iranian toll rate of $2 per barrel of oil, it would not distort trade or hurt the global economy too much.
Does Iran get to continue its nuclear enrichment program? While the U.S. would prefer a hard “no,” the most likely answer is that Iran will be permitted limited enrichment under careful international oversight – not dissimilar to the arrangement from 2015 to 2021. Iran has presumably been knocked some distance backward in its nuclear weapon aspirations by the recent bombing campaign.
Deal timing
Despite recent fierce rhetoric, it remains reasonably likely that the existing ceasefire holds. Betting markets point to a 66% chance that the ceasefire remains in effect through the end of the conflict (see the left side of the next chart). Even if there is a temporary resurgence of violence, betting markets assign an 83% likelihood that the war will be over by June 30.
Prediction markets show rising odds of conflict resolution but slow return to normal in Strait of Hormuz
As of April 13, 2026. Conflict end based on continuous 14-day period (beginning on/before specified date) without qualifying military action. Strait of Hormuz traffic returning to normal defined as averaging >60 ships per day. Sources: Kalshi, Polymarket, Bloomberg, RBC GAM
We have been slight optimists throughout this war, imagining that it would resolve slightly sooner than the market thought, and do marginally less economic and market damage. At times that has been a precarious prediction. It has felt on somewhat firmer ground over the past week, despite last weekend’s failure to reach the finish line. We remain somewhat optimistic.
Of course, even when the war is enduringly resolved, it will take time for energy markets to normalize. The right side of the prior chart reflects this: betting markets still show just a 53% chance that the flow of ships through the Strait of Hormuz will fully normalize by July. Of course, this leaves the possibility it returns to, say, 80% of normal well before that.
In general, though, it is reasonable to expect the supply of energy to normalize less quickly than it was shut off. It is difficult and time-consuming to restart halted wells. There has been some damage to infrastructure (though perhaps less than feared, as discussed in the “Energy infrastructure damage” section later). It takes time to get ships to where they need to be to load the energy and it takes even more time to get them to their destinations.
With the expectation that oil prices don’t fully revert to the pre-war norm even after a period of many months (see next chart), a geopolitical risk premium seems appropriate in this dangerous world. It could take time to de-mine parts of the Strait of Hormuz. Shipping insurance will likely be enduringly higher. Countries that drew down their energy reserves may want to rebuild them – temporarily enlarging demand.
The oil futures curve has shifted higher
As at April 10, 2026. Sources: Bloomberg, RBC GAM
-EL
War economics
The main economic message is that if the war with Iran is indeed over or ends fairly soon, the economic damage is set to be smaller than feared.
Recession risk down
While jittery, the betting-market probability of a U.S. recession in 2026 has declined from a peak of 40% to 32% today. Although it remains choppy, it has rebounded somewhat in recent days. In our view, the true recession risk may be even lower than this, for a few reasons. The U.S. is a net energy exporter. This shock could reverse quite quickly. The economy is benefiting from tailwinds including earlier rate cuts, ongoing fiscal stimulus and various AI-related drivers.
Odds of U.S. falling into a recession in 2026 are off their peak
As of April 13, 2026. Sources: Kalshi, Bloomberg, RBC GAM
Pessimists can note that the increase in U.S. gas spending has eaten about half of the US$18.3 billion in above-normal tax refunds that were supposed to drive the U.S. economy in the first half of 2026. Optimists can note the very same thing: the refunds remain more than capable of offsetting the pain of higher gas prices.
Current economic conditions hold up
Consistent with the idea that the recession risk is relatively tame, recent economic indicators have been solid. Global purchasing manager indices for manufacturers have actually advanced in early 2026 and proved surprisingly resilient as the war got underway in March (see next chart).
Manufacturing activity improved in most developed countries
As of March 2026. PMI refers to Purchasing Managers’ Index for manufacturing sector, a measure for economic activity. Sources: Haver Analytics, RBC GAM
Weekly U.S. initial jobless claims remain quite low right through early April (see next chart). This suggests little ill effect for the U.S. economy from high oil prices so far. Consistent with this, U.S. monthly payrolls for March managed a large 178,000 net new jobs – a good outcome even after adjusting for the end of a major health sector strike.
U.S. jobless claims still look fine
As of the week ending 04/04/2026. Sources: U.S. Department of Labor, Macrobond, RBC GAM
Inflation effect now visible
Unavoidably, higher oil and natural gas prices have nevertheless bled through into consumer prices. The U.S. Consumer Price index (CPI) just recorded a large (though expected) +0.9% month-over-month price increase for March. The State Street PriceStats real-time inflation index argues that April numbers should reveal a further significant leap (see next chart).
U.S. daily PriceStats inflation index suggests significant leap coming in April
State Street PriceStats Inflation Index as of April 07, 2026. Consumer Price Index (CPI) as of March 2026. Sources: State Street Global Markets Research, RBC GAM
But it remains important to recognize that if the war is indeed only temporary, then there should be room for several months of outright deflation later in the year as this distortion goes away. Year-over-year inflation at the end of 2026 may not be greatly higher than what was expected at the start of the year.
Central banks are unlikely to act with conviction unless they get the sense that a) the war will drag on for significantly longer and/or b) there is substantial passthrough from energy costs into other products.
When does the damage get harder to reverse?
Here are several ways of gauging whether the damage from the oil shock is becoming harder to reverse.
At the firm level, there are now anecdotal reports of some companies passing along a portion of their higher transportation costs to their retailer clients. That threatens to broaden consumer price pressures beyond purely energy-related products, though we think it is fairly limited for the moment. It merits close watching.
Fortunately, there are three factors that continue to argue that enduring damage should be limited.
We should not assume that the oil shock of 2022 and subsequent sustained inflation explosion provides a useful template for today. As we wrote in the March 10 MacroMemo, there are important differences:
“There are obvious parallels to the war between Russia and Ukraine, as in March 2022 that event also triggered a spike in oil prices to around $120 per barrel. While prices remained volatile and quite elevated for several months, oil prices ended the year at about $80 per barrel – approximately where they had started the year. This happened even though the war in Ukraine did not end.
While risk assets were quite unhappy during the first half of 2022, this had more to do with rapidly tightening monetary policy and spiking inflation. This inflation was partially a function of oil prices, but more so the result of overly strong post-pandemic demand and gnarled supply chains.”
The economic rule of thumb is that spillover effects should be limited until an energy shock has lasted for 3-6 months, at which point the danger starts to grow. This war is currently just 1.5 months old and has a good chance of being resolved before that danger zone is entered.
While short-term inflation expectations in the bond market have increased, this has lately partially unwound. Even more reassuringly, medium-term (5-year) inflation expectations have barely increased. Long-term (years 6-10) inflation expectations have actually fallen (see next chart).
U.S. short- and medium-term inflation expectations rise as oil prices surge
Note: As of April 10, 2026. Source: Bloomberg, RBC GAM
International oil dynamics
It remains the case that Asian and European economies are more adversely affected by the energy shock than North American ones.
While China is the largest single importer of energy from the Strait of Hormuz, it has been less badly affected than many Asian peers. This is for a few reasons.
China actually has relatively diverse hydrocarbon procurement, with significant domestic production and also substantial imports from other parts of the world.
Up until this new tentative U.S. blockade, China was still receiving the bulk of Iran’s exported oil despite the war.
China has a large strategic petroleum reserve, allowing it to replace a whopping 120 days of imports.
China normally refines and then exports approximately 1.2 to 1.5 million barrels of oil per day for its Asian neighbours. It is thought to be withholding a significant fraction of that during this energy shock, easing the extent to which China is pinched.
Conversely, many other Asian nations lack these protections and are made worse off by reduced Chinese refined product exports. The Philippines is at the top of the list, with Pakistan, Indonesia and Vietnam also prominently exposed to the shock.
In Europe, the immediate impact of higher energy prices is also quite large. However, it has been marginally cushioned by the fact that 13 of 21 countries (as per RBC Capital Markets) have instituted some form of policy response that effectively lowers the cost of energy – such as temporary tax cuts on fuel. Spain has been a disproportionate actor.
-EL
Energy infrastructure damage
While the recent U.S.-Iran ceasefire is encouraging, the nearly 6-week long conflict has already caused significant damage to energy and other civilian infrastructure in the region. As of March 23, the International Energy Agency said more than 40 energy assets across 9 countries in the Middle East had been “severely or very severely damaged.” This threatens to delay the normalization of supply chains even if a sustained ceasefire helps restore trade flows through the Strait of Hormuz.
That said, most incidents don’t seem to be associated with sustained production outages. In our view a significant if incomplete recovery in production appears likely in the coming months if the ceasefire holds.
As of April 6, Bloomberg’s tracking of energy infrastructure damage included the following:
9 oil refineries sustained damage and/or temporarily halted operations, including Ras Tanura in Saudi Arabia (the country’s largest refinery), Ruwais in the UAE (one of the world’s biggest refineries) and Sitra in Bahrain (the country’s only refinery). Those 3 facilities have a combined capacity of 1.8 million barrels per day (mb/d), or nearly 2% of global supply.
9 ports faced periodic disruptions, including Yanbu in Saudi Arabia (the terminus of the country’s East-West pipeline) and Fujairah in the UAE (the terminus of another pipeline that bypasses the Strait of Hormuz).
6 gas facilities sustained damage, including Qatar’s Ras Laffan Liquified Natural Gas (LNG) operation (the world’s largest liquefaction plant) and Iran’s South Pars gas field (the worlds’ largest gas field). QatarEnergy says significant damage to 2 of the 14 liquefaction trains at Ras Laffan – 17% of the facility’s capacity and 3-4% of global LNG production – could take 3-5 years to repair.
3 aluminum plants were damaged, including one of the world’s largest smelters in Al Taweelah, Abu Dhabi.
2 oil fields in Iraq and Saudi Arabia were reportedly targeted although details on the former are limited and no damage was reported to the latter.
1 nuclear plant in Iran (Bushehr) was reportedly targeted although the facility was undamaged.
In addition, a pumping station along Saudi Arabia’s East-West pipeline was reportedly struck by a drone shortly after the ceasefire announcement. That pipeline operated with 2-4 mb/d of spare capacity pre-conflict. More recently, it has been running at its full capacity of 7 mb/d, acting as a key mitigant to disruptions in the Strait of Hormuz.
The U.S. also struck military targets on Kharg Island, Iran’s main oil export hub, but did not target oil facilities. Iran still managed to export 1.6 mb/d of oil in March despite the conflict.
This is a long list of impacted facilities. But aside from Qatar’s Ras Laffan LNG plant, we haven’t heard of extended timelines to restore operations at other facilities. Rystad Energy calls that facility a clear outlier in terms of repair costs and timeframe.
Elsewhere, restoring production at Iran’s South Pars gas field may also prove a significant effort. Bahrain’s Sitra refinery is also near the top of the list in terms of disruption intensity.
But most of the affected facilities face less severe damage and shorter repair times (measured in months to quarters rather than years).
Overall, Rystad estimates a $25 billion repair bill associated with energy infrastructure damage. More significant disruption to gas fields and LNG facilities suggests a slower normalization compared with the oil market. That seems consistent with a more sustained premium in European natural gas and Asian LNG futures prices relative to crude oil (see chart).
Futures prices point to more prolonged disruption in gas/LNG markets
As at April 9, 2026. Sources: Bloomberg, RBC GAM
Beyond damage to energy infrastructure, lack of egress and limited storage capacity have forced output cuts that will take time to fully reverse. Middle Eastern producers reduced crude oil and condensate production by nearly 11.5 mb/d in March. This represents a more than 10% reduction in global supply (see chart below).
Some producers were able to simply slow output, which reduces complications associated with restoring production. But in cases where wells were fully shut in, returning to full production will be a slower and potentially incomplete process.
Overall, experts suggest it could take up to two months for output to return to pre-conflict levels if a ceasefire is sustained and the Strait is reopened. We think the 16% premium priced into crude oil futures by year end (see chart above) is more a reflection of tighter inventory levels and restocking and an ongoing geopolitical risk premium rather than significant supply disruptions that last into 2027.
Middle Eastern producers cut output by ~11.5 mb/d (>10% of global production)
As at April 2, 2026. Sources: Le Monde, Kpler, International Energy Agency (IEA), RBC GAM
-JN
A look at earnings through the fog of war
The outbreak of war in Iran rattled financial markets and spooked investors as spiking oil prices stoked fears of economic hardship. Stocks in many regions suffered declines near or exceeding 10% from their recent peaks (see next chart). Investor sentiment plunged to extreme pessimism as the war intensified (see subsequent chart).
Major equity market indices suffered declines near or exceeding 10% from recent peaks
As of April 10, 2026. Sources: Bloomberg, RBC GAM
Investor sentiment plunged to extreme pessimism by early April
As of April 9, 2026. Sources: Ned Davis Research Daily Trading Sentiment Composite, RBC GAM
Since then, though, most markets have enjoyed a meaningful recovery as discussions toward a ceasefire have progressed in fits and starts – diminishing fears of a worst-case scenario. Volatility may persist as the geopolitical situation continues to evolve. But investors should be comforted by the fact that macro fundamentals were in solid shape leading into the war, and furthermore that corporate profit estimates have actually been rising rather than falling since the onset of the conflict.
In the past month, analysts’ earnings estimates for the S&P 500 were revised higher. This is the continuation of a trend that has been in place for the past year (see next chart). The projections have gone up in each of the past 12 months, with the March 2026 increase the largest upgrade in the sequence.
2026 earnings estimates for S&P 500 continue to rise
As of April 10, 2026. Sources: Bloomberg, RBC GAM
While counterintuitive, an important motivation for the upgrade was likely that oil prices actually act as a tailwind to corporate profits. Based on a historical analysis by Goldman Sachs, every 10% increase in the price of brent crude oil translates to an approximate 0.4% boost to S&P 500 earnings per share (EPS) (see next table). The 50% increase in oil prices from the start of the year should therefore be about 2.0% to the EPS growth estimate, all else equal.
How changing macro variables affect S&P 500 earnings per shares
As of March 2026. Source: Goldman Sachs Global Investment Research
There are, of course, other moving parts, as outlined in the table. A stronger U.S. dollar weighs on corporate profits, while the increase in bond yields is another tailwind. But the changes in these variables since the start of the year have been relatively small and inconsequential.
The most impactful variable is U.S. gross domestic product (GDP), with a sensitivity of 3.4% to S&P 500 EPS for every 100 basis-point-change in the economic growth rate. As a result, a negative scenario in which sustainably high oil prices tip the economy into recession would likely dominate the fundamental picture. This could cause an outright contraction in corporate profits. So far, though, our base case is that the oil price shock that we have seen is not severe enough to pull the U.S. economy into recession.
Energy has been dominating headlines due to the war in the Middle East, but it’s important not to lose sight of other forces also at work in the earnings revisions. The massive spending related to artificial intelligence (AI) continues to fuel profit growth in the S&P 500’s largest sector. Since the war began, S&P 500 profit estimates for 2026 have been revised up by 2.3%. Half of that increase is attributable to higher energy prices, and the other half to upgrades to Information Technology profit estimates (see next table).
Energy and Information Technology drive rising estimates of S&P 500 earnings per share
As of April 9, 2026. Sources: Bloomberg, RBC GAM
Digging into the details, energy profit estimates have risen a substantial 34.5% since the war began. That translates to only $3.86 per share of extra S&P 500 earnings as Energy is a relatively small part of the S&P 500 profit picture. In comparison, Information Technology profit estimates have increased by a smaller 4.68%. But because of the sector’s far larger weight in the earnings pool, that change adds a sizeable $3.66 per share to the S&P 500 profit pool. Other sectors’ estimates were relatively unchanged over the same period.
Gazing ahead to 2027, earnings estimates for next year for the Energy sector are up by 14.4% since the start of the war, adding $1.94 EPS to the S&P 500. Technology earnings are also rising for 2027, helped by significant investment in AI-related infrastructure and computing capabilities. Since the start of the war alone, earnings estimates for Information Technology in 2027 have been revised higher by 9.53% or a monstrous $9.51/share. These figures bring the EPS growth estimate for 2026 and 2027 to 18.1% and 18.4%, respectively, with Information Technology accounting for half of the entire index growth in both years.
Further supporting the idea that the earnings outlook remains solid is the fact that investment-grade credit spreads have correlated closely to earnings growth since the war began. In the past, investment-grade credit spreads have correlated closely to earnings growth (see next chart). Significant widening in credit spreads often foreshadows challenges that will hinder corporate profit growth.
But since the war began, credit spreads only widened around 20 basis points. This pales in comparison to the typical widening in excess of 100 basis points that preceded prior meaningful profit slowdowns. The fact that spreads have tightened to near historic levels perhaps reflects complacency among investors. But it also suggests an optimistic view that the geopolitical conflict in the Middle East is likely to be short-lived and/or that any negative consequences should be limited.
Investment-grade credit spreads have correlated closely to earnings growth
As of April 9, 2026. Sources: Bloomberg, RBC GAM
Integrating all of this information into our scenario analysis for the S&P 500 reveals a variety of possible paths ahead. The following chart highlights three scenarios (bull, base and bear case) based on various earnings projections and valuation assumptions. While the S&P 500 has been tracking near or above the bull-case scenario for the past year or so, the trajectory for all three scenarios has lately inflected higher to reflect analysts’ earnings forecast upgrades. This, combined with the pullback in stocks over the past month, has boosted the return potential for stocks from here, even if the subsequent rebound has eaten into that return potential somewhat.
Stocks remain volatile and the bear case should not be entirely disregarded. However, the refreshed base-case outlook remains reasonable: a mid-single digit return for the S&P 500 through the end of 2027 in an environment in which the economy continues to grow, inflation pressures prove temporary and corporate profits rise at a rapid pace – fanned in part by the spurt of higher oil prices, but mostly by continued strength in the technology sector.
Our scenario analysis for S&P 500 focuses on bull, base and bear cases
As of April 10, 2026. Sources: Bloomberg, RBC GAM
-ES
AI chip obsolescence concerns
U.S. tech companies continue to ramp up investment in AI data centres. The consensus estimate for 2026 CapEx by the big 5 hyperscalers has risen to $677 billion from $536 billion at the start of the year (see chart). That would represent a 63% increase relative to 2025, which itself was up 73% from a year earlier. This eye-watering pace of growth and the sheer scale of spending – $677 billion amounts to more than 2% of U.S. GDP – continues to generate concerns about over-investment.
Hyperscaler CapEx estimates continue to be revised higher
Note: As at April 2, 2026. Estimated capex by AMZN, GOOGL, META, MSFT and ORCL. Source: Bloomberg, RBC GAM
At a micro level, data centre economics hinge on revenue drivers (price charged per GPU-hour and utilization rates) and cost structure.
From the revenue side, much therefore depends on the extent to which AI models become ever-more-useful – a subject examined in our last MacroMemo, with broadly promising conclusions.
From the expenditure side, servers account for roughly half of data centre costs. The price of GPUs and how quickly they need to be replaced are thus key issues. The latter has drawn more scrutiny recently given two seemingly contradictory trends: rapidly improving GPU performance and shorter product cycles that increase obsolescence risk for the chips, versus tech companies that are lengthening their assumptions for the economically useful life of the chips they purchase.
On the first point, cutting edge GPUs are becoming 37% more powerful every year, on average, when adjusted for price (see chart). For instance, Nvidia’s B100 Blackwell chips saw a 77% performance jump (with no increase in power draw) relative to the company’s previous generation H100 Hopper chips released two years earlier – and they only cost 4% more. Wolfe Research estimates the internal rate of return on a data centre using the newer GPUs would be about two-thirds more than a facility using the previous generation chips.
Performance/price ratio improves
As at March 30, 2026. FLOP is floating-point operations per second, a measure of computing performance. Sources: Epoch AI, RBC GAM
Cloud service providers (CSPs) using the latest generation hardware are able to charge premium prices while the cost to rent out previous generation chips tends to fall when newer technology is released (see chart below). That can deal a blow to CSPs with older chips that are renewing enterprise contracts (capacity tends to be sold in 1- to 3-year agreements) or selling excess capacity on the spot market. As Nvidia and other chipmakers shift to annual product cycles from semi-annual releases previously, this obsolescence risk could intensify.
Inference prices take a hit when a new generation of Nvidia chips is released
As at April 7, 2026. Sources: Silicon Data, Bloomberg, RBC GAM
Despite this, many hyperscalers have extended useful life assumptions for their servers and network equipment by as much as 3 years since 2020 (see table below). Critics suggest overly optimistic useful life assumptions are designed to reduce annual depreciation expenses and boost profits. Michael Burry of The Big Short fame suggests hyperscalers will understate depreciation by $176 billion between 2026 and 2028.
Most hyperscalers extend useful life assumptions for their servers and network equipment
As at December 8, 2025. In 2025, AMZN shortened useful life to 5 years from 6 for some equipment. Sources: Michael Burry, U.S. Security and Exchange Commission (SEC) company filings, RBC GAM
Longer depreciation schedules combined with the rapid pace of hardware innovation raise the risk that overvalued, obsolete assets eventually have to be written down, with profits ultimately taking a hit. A data centre that was expected to be profitable based on an assumed 6-year useful life for its chips might turn out to be a losing proposition if its GPUs have to be replaced after 4 years.
Possibly representing a change of heart, Amazon recently reduced the assumed useful life of some of its equipment to 5 years from 6 previously, citing “the increased pace of technology development, particularly in the area of artificial intelligence and machine learning.”
But while previous generation chips might not be able to train the most advanced AI models or command premium rental pricing, they can still be used for high-end inferencing and eventually less demanding tasks and analytics. For instance, OpenAI still uses Nvidia’s A100 GPUs (released in 2020) for inference and CoreWeave said its own A100 chips remain fully booked. CoreWeave was apparently able to renew contracts based on its H100 chips (released in 2022 and still sold by Nvidia) at 95% of their original rental prices.
As the AI boom is relatively nascent – Nvidia only released its first AI-focused GPUs in 2018 and ChatGPT’s first commercial release was just over 3 years ago – there is limited historical reference to inform useful life assumptions. Whether tech companies have become too optimistic in their useful life assumptions will depend on how a number of factors evolve in the coming years:
Supply vs. demand: Between Nvidia’s reported $1 trillion order book and CSPs’ growing enterprise backlogs, compute is clearly at a premium. Inference prices are the costs of running data through a trained AI model to generate outputs. These costs have been rising year-to-date, even for previous generation chips (see chart above). This allows CSPs to continue using older hardware and rebook contracts at only modestly lower prices. But if new data centre capacity coming online outpaces AI adoption, the balance between supply and demand would shift and existing hardware could lose pricing power.
Pace of hardware improvement: If the 37% average annual improvement in the performance-to-price ratio of GPUs accelerates, previous generation chips could see their inference pricing fall faster. They could also become economically obsolete sooner. Oppositely, a slower pace of technology improvement – perhaps due to physical constraints on semiconductor fabrication – would help older GPUs maintain their value and pricing power.
Training vs. inference: While training previously accounted for most AI workloads, that balance will shift in the coming years. McKinsey sees AI inference demand matching training workloads this year and accounting for 1.5x as much data centre demand by 2030 (see chart below). With older chips generally remaining useful for inference, that should support longer economic usefulness.
Physical life: Much of the focus is on the (typically shorter) economically useful life of chips – how long they’re able to generate a sufficient return – rather than their physical life. But chips can fail and wear out and might need to be replaced before their expected economically useful life. A 2024 Meta study found a 9% (annualized) failure rate of H100 GPUs during Llama 3 model training. Hardware utilization rates and other factors can impact physical lifespan, but further advances in heat management and cooling technology and improving quality control could help to broadly reduce failure rates.
AI inference is expected to account for a larger share of data centre workloads
As at April 7, 2026. Sources: McKinsey & Company, RBC GAM
In our view, it’s not clear that longer economically useful life assumptions are simply an accounting gimmick. There appear to be good reasons why chips can generate revenue for longer, even if they aren’t able to do the most intensive model training and inference for years on end. But given how sensitive data centre returns might be to a shorter useful life, some degree of skepticism is healthy, and keeping an eye on the aforementioned trends seems prudent.
-JN
Canada as resource superpower
Canada is internationally recognized as a resource-rich nation – a long-standing point of strength for the country.
This bounty is arguably becoming even more valuable as the world becomes more volatile and dangerous. With the disruption to the flow of energy and other commodities through the Strait of Hormuz as just the latest motivation, countries are increasingly focused on procuring their food, energy and other critical resources from stable, reliable countries like Canada. In a multipolar world, Canada is an attractive source of raw inputs for the U.S. and China alike.
What follows are a series of charts and tables that highlight Canada’s remarkable pre-eminence across a wide range of resource types.
Canada is the fourth largest oil producer in the world (see next chart), and the fifth largest natural gas producer (see subsequent chart). Based on reserve data, there is room for significant growth in both.
Canada is fourth in global oil production, after the U.S., Saudi Arabia and Russia
As of December 2025. Production of crude oil including lease condensate. Sources: Energy Information Administration (EIA), Macrobond, RBC GAM
U.S. leads top natural gas producing countries; Canada is fifth
Sources: The Energy Institute Statistical Review of World Energy 2025, RBC
Canada is ranked first in global potash production (see table below) and fifth in overall fertilizer production (as per the chart after that). It is also second in uranium production at a time when nuclear energy is coming back into vogue. Canada is third and fourth in diamond and gold production, respectively. The latter has achieved renewed global interest as a substitute for the U.S. dollar.
Canada leads in global potash production, top three in uranium
Sources: Natural Resources Canada, RBC GAM
Canada is fifth in global fertilizer production, just after U.S.
Production of nitrogen, phosphate and potassium fertilizers. Sources: Foreign Agricultural Service, U.S. Department of Agriculture (USDA), RBC GAM
In the metals space, and as per the earlier table, Canada is fourth in global nickel production, seventh in lithium and cobalt, and eighth in iron ore. Canada is also fourth globally in aluminum production (see next chart).
Canada is fourth globally in aluminum production
Sources: Natural Resources Canada, RBC GAM
Meanwhile, the country is only slightly less remarkable in its agriculture prowess. It ranks sixth globally in wheat production, eight in soybeans and tenth in corn (see next three charts).
Canada ranks sixth in global wheat production, just after U.S.
Sources: USDA, Macrobond, RBC GAM
U.S. ranks second in global soybean production, Canada ranks eighth
Sources: USDA, Macrobond, RBC GAM
U.S. ranks first in global corn production, Canada ranks 10th
Sources: USDA, Macrobond, RBC GAM
As it happens, after a decade of political orientation toward matters of social policy, the Canadian government is again focused on growing the country’s economy, raising productivity and enabling large infrastructure and resource projects.
The current federal government appears to be on the cusp of attaining majority status after a number of defections from other parties and with a series of imminent by-elections as this goes to print.
With a majority in Parliament, a federal government focused on growing the economy could prove a potent combination for increasing Canada’s resource production significantly in the years ahead.
-EL & VL