~With contributions from Vivien Lee, Aaron Ma and Eric Savoie
Iran war update
Ones interpretation of the current state of the U.S.-Iran conflict depends critically on the area of focus.
The good news
Arguing for quite an optimistic interpretation, the war itself has been on pause for nearly a month (see next chart). Missiles and drones are no longer whistling through the air toward their targets. This is certainly very good news. For the great majority of wars, this would be just about the only thing that mattered.
Daily count of missiles and drones launched by Iran remains low
As of 05/01/2026. Data compiled by RBC AI tool web scan of news reports, third-party analyses and ministry of defence official statements from Gulf countries. Source: RBC GAM
The mixed news
Arguing for a mixed interpretation, both sides continue to talk but have not yet secured a peace accord.
At various points, there had seemed to be a mutual agreement over the basic framework for a deal, if not the specific contours of each point. But recent comments suggest the two sides have drifted somewhat further apart. Various plans to talk have been cancelled and tensions have increased. On the other hand, Pakistan continues to serve as mediator and behind the scenes China is likely prodding Iran toward a deal.
The key remaining sticking points are over the control of the Strait of Hormuz and Iranian nuclear refinement. It is unclear what exactly the solution will be, but we continue to believe there is a path toward a deal, and that both parties are strongly incented to reach one.
On Hormuz, the U.S. could hold firm on insisting upon open access, or it could hold its nose and accept an Iranian fee or joint Iran-Oman administrative control. On nuclear refinement, Iran had within the past month suggested a willingness to halt its nuclear enrichment program for five years, and the country is not so many years removed from having accepted intensive on-the-ground international monitoring.
The bad news
Finally, arguing for a negative interpretation, the Strait of Hormuz remains mostly closed (see next chart). This is obviously a big problem for the global oil and natural gas market. It explains why oil prices remain extremely elevated and is the most relevant variable for the global economy.
Ship crossings in Strait of Hormuz collapsed but reviving
As of 05/01/2026. Sources: Bloomberg, RBC GAM
The U.S. recently announced Project Freedom, which is intended to escort ships through the Strait of Hormuz starting on May 4. But Iran has threatened to escalate the conflict if that actually happens and shipowners have indicated they are unclear how the U.S. plan would work. They are reluctant to steam through the Strait of Hormuz even under notional U.S. protection. At a minimum, it is not a long-term solution.
Taken altogether, the conflict is in a much better place than it was in March, but the basic shortage of fossil fuels is no closer to being resolved. Betting markets have become notably glummer over the past week, shifting from pricing a 61% chance that the U.S. blockade is lifted by the end of May to just 31% now (see next chart). Whereas a week ago shipping through the Strait of Hormuz was deemed likely to completely normalize by the end of June, that probability is now down to 45%.
We stubbornly remain slight optimists – the odds of the blockade being lifted by May 31 seem better than 31% to us. But a near-term resolution is not certain.
Prediction markets are becoming more pessimistic about Strait of Hormuz normalization
As at 05/04/2026. Strait of Hormuz normalization = 7-day average of Strait transit calls (arrival of ships) is 60 or greater by specified data. Sources: Polymarket, RBC GAM
Stock markets arguably overreacted in late March when they fell by roughly 10% in response to what is fundamentally a temporary energy shock. However, the record highs several have achieved in recent weeks make for a less attractive tactical entry point.
In fairness, the stock market’s strength has been about far more than downplaying the energy shock. Tech sector earnings are genuinely soaring and optimism is high about an AI-enabled future.
Still, there is now again a clear downside risk emanating from the energy space, related to the following issues:
A deal could prove elusive for longer than expected.
So long as Hormuz is closed, oil prices will likely rise over time as shortages intensify, rather than simply remain high.
Some of the anticipated economic and inflation damage will show up later and potentially dim sentiment.
The risk is mounting of second-order inflation effects now that the energy shock is in its third month.
Whenever the blockades are lifted, it will take several months to normalize the energy supply chain.
“Normal” energy prices will likely be higher after the war than before the war.
Government fiscal positions are being damaged by the war due to widespread energy subsidies (see the next discussion).
Government support
In assessing the relative inflation and economic damage that this energy shock inflicts upon various countries, it is not enough just to observe that the biggest crude oil and wholesale natural gas price increases have been recorded in Asia and Europe.
Yes, this argues that, all else equal, those markets should suffer the greatest negative blow. But governments have been scrambling to deflect some of the pain (see next table).
How governments are responding to rising fuel costs
As at 05/01/2026. Sources: International Energy Agency (IEA), RBC GAM
Prominently, quite a range of countries have implemented programs to ease the pain of consumers. Many countries have cut their fuel tax, including Canada, Germany, Spain, Australia, Brazil, Turkiye, India and Vietnam. Meanwhile, Indonesia and Thailand have introduced fuel subsidies – a different way of reaching the same goal.
The strongest actions have been taken in a handful of countries that have outright capped their retail fuel price, including China, Japan, South Korea, Mexico and Poland. The Philippines has also been supportive via a variety of measures.
Overall, Asian and European countries have been particularly intensive in their consumer supports, helping to reduce the impact on their inflation.
The U.S. is an interesting exception. Although it is more buffered than most countries by virtue of its status as a net energy exporter, it has allowed the entirety of the energy shock to reach its consumers. More obscurely, the fact that the U.S. has a lower gas tax to begin with means that the percent swing in consumer-facing gas prices is larger in the U.S. than in an equivalent market with a higher gas tax (because such taxes are usually a flat rate).
The conclusion is not that the U.S. Consumer Price Index (CPI) will be the most adversely affected CPI, but that its advantage versus other markets may be smaller than generally anticipated. The U.K. has also been relatively constrained in its support actions.
Overall, Asian and European countries have been particularly intensive in their consumer supports, helping to reduce the impact on their inflation.
Some countries – notably Asian markets confronting literal physical shortages of fossil fuels – have also made efforts to discourage consumption. They have adopted a variety of measures such as encouraging people to work from home and driving restrictions (see the rightmost column in the above table). This doesn’t so much eliminate the economic pain as redirect it in a way that hopefully reduces the scale of the damage.
-EL
Shifting monetary regime
In the U.S., exiting Fed Chair Powell held his last press conference as chair, but will remain on the Federal Open Market Committee (FOMC) as a governor. The White House recently dropped criminal charges against him, clearing the way for Republican senators to support the confirmation of the nominee for Fed Chair, Kevin Warsh. This is likely to happen in the coming weeks.
Warsh seems likely to prove reasonably independent from White House pressure. He will likely show a general preference for a smaller Fed balance sheet and a low policy rate – with the potential for a slightly steeper yield curve.
Our central bank tracker argues that the era of global monetary policy easing has effectively ended (see next chart). It spanned 2024 to early 2026. These periods tend to last a few years, and the best guess is that a period of modest to moderate rate hiking may come next.
Era of interest rate cuts has faded
As of 04/30/2026. Based on policy rates for 30 countries. Sources: Haver Analytics, RBC GAM
A pivot to tightening is not automatic – see the long period of relative stability spanning the mid-2010s. But market expectations now tilt toward rate hikes across a number of major developed markets (see next chart). Additionally, the Reserve Bank of Australia – sometimes a bellwether – already raised rates this spring and appears on track to do so again in the near future.
Markets are starting to expect rate hikes across G7 central banks
As of 04/30/2026. Dotted lines indicate futures pricing. Sources: Bloomberg, RBC GAM
Our own refreshed views now broadly align with this bias, with the Fed expected to keep its fed funds rate unchanged, and unhurried monetary tightening from the ECB, Bank of England, Bank of Japan and Bank of Canada.
It makes sense that central banks are starting to think about rate hikes. Several have slightly stimulative rates and yet economic growth is neutral and inflation is elevated.
Developed-world manufacturing Purchasing Managers’ Indices (PMIs) have unexpectedly accelerated in recent months, defying the energy shock (see next chart). Quite a range of economic indicators have lately been coming in positive.
Manufacturing activity has improved in most developed countries
As of April 2026. PMI refers to Purchasing Managers’ Index for manufacturing sector, a measure for economic activity. Sources: Haver Analytics, RBC GAM
We continue to see important tailwinds for 2026 growth (see next table). Note that while monetary policy may start to tighten, the legacy of years of rate cuts still provides a helping hand to growth across 2026.
Growth tailwinds still exist for 2026
As of 03/04/2026. Source: RBC GAM
Meanwhile, on the inflation side, while central banks can look through the direct temporary effect of the energy shock, the scope for spillover is rising now that the shock has lasted multiple months. Oil prices are not expected to fully normalize even after the Strait of Hormuz is re-opened.
What should investors know as we edge toward a rate-tightening regime? Returning to the two-by-two matrix we have used in the past, we can gauge expected stock and bond market returns (see next table). The environment is seemingly shifting from the top-left quadrant (“no recession” plus monetary “easing”) to the bottom-left quadrant (“no recession” plus monetary “tightening”). With the important caveat that each cycle is ultimately unique, this regime is frequently associated with solid but less impressive stock market gains than the prior one, and notably less impressive fixed income returns. It is still the second-best regime to be in out of the four, but less attractive overall.
Monetary policy and economic regime matrix helps gauge market returns
As of 04/28/2026. Daily S&P 500 data begins in 1955 and U.S. 10-year bond data begins in 1962. Return and standard deviation are annualized. Min and Max are of the cumulative returns for continuous days in the regime. The % of time figures don’t add up to 100% because there are some periods where monetary policy is neither tightening nor easing. These periods were excluded from the analysis. Source: RBC GAM
-EL
AI excitement fuels semiconductor boom
The war in Iran and higher oil prices remain a challenge to economic activity in the near term. But history suggests that financial markets tend to look past near-term challenges to focus on long-term opportunities. The powerful recovery in stocks since the March sell-off likely reflects investors shifting their sights to a future where the Middle East conflict eases, oil prices stabilize and artificial intelligence (AI) spending continues to fan corporate profits.
The technology sector, in particular, displayed marked leadership in the latest stock-market rally, but with a critical nuance. Software stocks continued to underperform, extending a trend that began in late 2025. In contrast, semiconductor stocks experienced explosive gains (see next chart).
While AI tools are disrupting software providers, the rising demand for computing capacity is benefiting companies that supply the ingredients for AI infrastructure. The surge in semiconductor stock prices suggests the future for AI-related investment is bright, and still brightening, but we also recognize that given the extent of recent gains and how much valuations have ballooned, semiconductor stocks would be vulnerable should AI’s promise falter.
Semiconductors, software and S&P 500 performance reflect confidence in AI’s future
As of 04/30/2026. Sources: Bloomberg, RBC GAM
It is worth drawing attention to the semiconductor industry because the outsized gains of stocks in that space have earned it a meaningful presence in the S&P 500 Index. As of late April, the semiconductor industry’s weighting in the cap-weighted S&P 500 Index was 16.7%, a figure that has tripled since 2022 and is nearly half of Technology’s 34.9% weighting (see next chart).
As a result of this increased weighting, semiconductor stocks can have meaningful influence on the performance of the S&P 500 either to the upside if positive momentum continues, or to the downside should the recent trend reverse.
Industry weights rising for S&P 500 technology sector and semiconductors, semiconductor equipment
As of 04/28/2026. Sources: Bloomberg, RBC GAM
By some technical measures, the recent rally in in the Philadelphia Semiconductor Index (SOX) is extremely stretched and vulnerable to a pull-back. The index has rallied 48% since late March, with a 150% rally over the past year. The index’s 14-week relative strength index (RSI), a measure of intensity of price change over a 14-week period, climbed above the 70-level indicating “overbought” (see next chart).
Moreover, the distance the SOX index lies above its 200-day moving average climbed more than 2 standard deviations above its 30-year average. The index is now at its highest reading since the late 1990s technology bubble (refer to subsequent chart).
These indicators suggest the speed and magnitude of the recent rally are at historic extremes that could succumb to the forces of mean-reversion. Sustaining such gains will likely require AI fundamentals to remain strong and for the outlook around AI investment to continue to strengthen.
Philadelphia Semiconductor Index (SOX) rally continues but is vulnerable to a pull-back
As of 05/01/2026. Sources: Bloomberg, RBC GAM
Philadelphia Semiconductor Index lies above 200-day moving average
As of 04/30/2026. Sources: Bloomberg, RBC GAM
That said, these recent gains may very well be warranted by the massive increase in capital spending plans by the AI hyperscalers. The next chart plots the annual capital expenditures of the Magnificent 7 companies since 2020, with forecasts for 2026 through 2028. Two critical observations are worth highlighting.
The first is that AI-related capital spending is expected to grow extremely fast from US$378B last year to US$668B this year with further increases in 2027 and 2028. But what may be even more striking is that these expected figures have been revised significantly higher since the beginning of the year. Compared to December 31, 2025, the estimated capital spending for 2026, 2027 and 2028 has risen by US$166B, US$231B and US$275B, respectively.
That is, the total expected capital spending over the next three years (2026 inclusive) has risen to US$2.31T. That’s a US$672B or 41% increase over the US$1.64T that was expected just four months ago.
Magnificent 7 capital expenditure estimates are rising quickly
As of 05/01/2026. Magnificent 7 includes Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. Sources: Bloomberg, RBC GAM
The increased spending has a real impact on earnings projections for corporations involved in the AI buildout, especially semiconductors. The biggest improvements in profit expectations so far this year have been in the following industry groups (see next chart):
semiconductors (+34.7%)
energy (+16.4%)
technology hardware & equipment (+12.5%)
materials (+11.4%).
Energy has been helped by higher oil prices stemming from the war in Iran, but the increase in semiconductor earnings is extreme and stands out. In fact, most industry groups have not seen a significant change in earnings expectations, so the revisions this year have been highly concentrated in segments affected by energy and artificial intelligence.
Profit expectations rising fastest for semi-conductors, energy, technology hardware and materials
As of 04/28/2026. Sources: Bloomberg, RBC GAM
In this environment, investors have been willing to pay quite a high price for semiconductor stocks. The price-to-sales ratio for the semiconductor industry has soared to 17.4x. That’s almost three times higher than its long-term average of 5.0x (see next chart).
The valuation is rising alongside a surge in sales growth, though it is worth acknowledging that, in the past, semiconductor sales growth has been highly cyclical. As a result, the extreme valuation may not be appropriately compensating investors for the risk of any slowdown.
S&P 500 semi-conductors and semi-conductor equipment stocks rising as sales surge
As of 04/29/2026. Sources: Bloomberg, RBC GAM
The high valuation in semiconductors stands out relative to its own history and especially compared to other industry groups, even those within the technology sector. The next chart plots the price-to-sales ratios for each industry group within the S&P 500, color coded by sector. Semiconductors, at 17.4x price-to-sales ratio, dwarfs the other technology industries – software & services and technology hardware & equipment – which are trading at 7.8x and 6.7x, respectively.
While the valuation readings for all these industry groups are above their respective long-term averages, that of semiconductors is an outlier. It reflects an expectation that exceptional and durable sales growth for that industry lies ahead.
S&P 500 price-to-sales ratios show high expectations for future semiconductor growth
As of 04/28/2026. Sources: Bloomberg, RBC GAM
It is becoming increasingly clear that investors have embraced the idea that AI infrastructure is becoming a key investment priority for the economy. The surge in capital spending and upward revisions to earnings expectations is evidence that the AI revolution is real and promising. But given the fact that semiconductor demand has historically featured cyclical patterns, and that valuations and technical signals are stretched, investors may want to temper expectations. Risk management in this environment is especially critical.
-ES
Will AI be a winner-take-all technology?
Given the exponential and accelerating pace of improvement in AI model performance, it’s worth contemplating whether the market might eventually be dominated by one or two big players whose ultra-capable foundation models are widely adopted across a variety of enterprise and consumer applications. Or, might there instead be space for many competitors, including smaller firms and more niche, task-specific models? Will it be winner-take-all or a diverse, competitive marketplace?
This is an important question for investors. It informs whether the correct investment strategy is to focus on first-movers, or instead upon a broader set of players.
The tech industry offers many examples of winner-take-all outcomes – monopolies in which one company emerges as the clear leader in a market segment. There’s Meta in social media, Google in search, Amazon in online retail. Tech lends itself to market concentration for a number of reasons:
Cost structure: Tech platforms typically require significant upfront investment in R&D, human capital and intellectual property. However, the marginal cost of increasing output or adding users is close to zero. High barriers to entry combined with significant economies of scale tend to drive market concentration.
Network effects: Technology often exhibits strong network effects, becoming more valuable with every additional user and helping the big get bigger. AI tools might allow a coder to build a Facebook clone with relative ease, but without 3 billion active monthly users it would have limited value.
Data network effects: More users generate more data – the “new gold” in an increasingly tech-centric economy. This allows for faster improvement in products and algorithms, attracting even more users to the platform in a positive feedback loop.
Mergers & acquisitions (M&A): Big tech companies actively buy up smaller players, increasing market concentration. The Centre for Research on Multinational Corporations found Alphabet, Amazon, Apple, Meta and Microsoft acquired at least 191 companies between 2019-2025 – one every 11 days, on average. Regulators blocked only two mergers.
Some of these factors are present in AI as well:
Massive investment in compute to train frontier models, plus the need to continuously reinvest to stay ahead of competitors, acts as a barrier to entry. OpenAI reportedly spent $5 billion on R&D compute in 2024. Only $400 million of this went toward the final training run of GPT-4.5. xAI also spent nearly $500 million to train Grok 4 – almost twice as much as Grok 3, which was released only 5 months earlier. Indeed, the cost of training frontier models has doubled every 7 months on average (see chart below).
User feedback and data generated during inference (making predictions using trained models) can help fine-tune models, allowing those with a larger user base to improve faster. And with models now enhancing themselves via recursive learning, more successful platforms should develop at an even faster pace, again supporting those with scale.
As models become more integrated into business processes, using in-context learning to improve output, barriers to switching platforms rise. This creates a competitive moat for more successful players and a challenging environment for new entrants. A 2025 survey by Menlo Ventures found only 11% of enterprises switched model vendors in the past year, while 66% of enterprises chose to upgrade models within their existing provider and 23% made no change.
M&A has been focused more on vertical integration across infrastructure, models and applications rather than horizontal acquisitions of models or coding agents. However, McKinsey expects to see more capability-driven AI buyouts going forward. Compiling its 2026 list of the top 50 AI companies, Forbes noted three from the previous year’s list were acquired or “acquihired” (buying a company mainly for its employees) by tech giants.
Cost of training frontier models is rising exponentially
As at 04/28/2026. Sources: Epoch AI, RBC GAM
Some see these dynamics eventually resulting in a winner-take-all outcome. The most advanced foundation model, trained on the best and largest dataset, will be adapted to a variety of use cases and become the dominant solution for enterprise and consumer applications.
The hyperscaling companies currently plowing as many resources as they can muster into their models seem to think that it is critical to be in the lead. This implies that they fear there will not be room for more than a few models once the market matures. Supporting this, to the extent AI models are becoming capable of enhancing themselves, the best models could start to improve at such a clip that there is no hope of the laggards ever catching up.
Yet none of the leading developers appear to be breaking away from the pack when it comes to performance. In fact, scores across frontier models are converging. According to Stanford’s latest AI Index Report, Arena reviewed the performance scores of the top 4 models (from Anthropic, xAI, Google and OpenAI). They differed by just 22 points (less than 2%) as of March 2026, compared with a 97-point gap a year earlier. Three years ago, OpenAI had a more than 200-point lead over Google.
To the extent these models are being built with the same computer chips and being trained on the same repository of internet information, perhaps the similarities should not be such a surprise.
DeepSeek and other open model developers like Meta and Mistral tend to lag behind closed proprietary models. But they are able to achieve 90% of the performance at just one-eighth the cost per token, according to an MIT study.
Some developers are taking more of a fast follower approach. They’re leveraging the work done by first movers and saving on R&D by avoiding dead ends in product development and having to buy expensive bleeding-edge semiconductors.
DeepSeek caused a stir in early 2025 with the release of its R1 model, which was reportedly trained for less than $6 million using the output of significantly more expensive models. DeepSeek and other open model developers like Meta and Mistral tend to lag behind closed proprietary models (see chart below). But they are able to achieve 90% of the performance at just one-eighth the cost per token, according to an MIT study.
Open-weight models only lag modestly in performance but cost much less
As at 05/01/2026. Sources: Epoch AI, RBC GAM
Open models, which publish some details like model weights and source code – offer other advantages like greater transparency, customization and data security. These might attract enterprises with proprietary data. Closed models currently account for about 80% of model usage, but if more users are willing to sacrifice some performance for significant cost savings and other benefits, open models could commoditize inference and drive token prices down toward the cost of compute.
Similarly, fast-following companies could manage to avoid some of the missteps of the first movers, arriving at the same finish line not much later and at a much lower cost. That competitive pressure would impair the economies of scale that might otherwise help one or more of the existing leading models from dominating the market.
It’s also not clear that the large language models (LLMs) many developers are focusing on will ultimately prove to be the most powerful AI technology or the right solution for every application. Yann LeCun, a top AI researcher and former chief AI scientist at Meta, thinks LLMs are a dead end and won’t reach human-level intelligence. Instead, he advocates world models that can predict how the world will respond to actions, processing different types of data (including audio, video, images and scientific data) rather than the vast amounts of text used to train LLMs.
Active asset management is a business in which one’s investment strategy must be superior to the other market participants. Having the most advanced model makes sense here, too.
Whether LLMs, world models or some other form of prediction becomes the most powerful AI tool, it seems plausible that different approaches will be favoured for different use cases. Even within LLMs, some experts suggest the competitive landscape will be split into many verticals that are tailored to the specific needs of different industries, rather than a single, powerful foundation model that is adapted to a variety of tasks and applications.
Fields such as pharmaceutical research require constant innovation that clearly benefits from having the smartest possible model. Active asset management is a business in which one’s investment strategy must be superior to the other market participants. Having the most advanced model makes sense here, too.
On the other hand, some tasks don’t require genius-level artificial intelligence. Many accounting and legal services require thoroughness and competence more than brilliance. Cheaper, more basic models may suffice.
While some of the dynamics that traditionally favour scale in tech are also present in AI, there are enough key differences that a winner-take-all outcome is not a foregone conclusion. It is perhaps not even the most likely outcome.
The new AI models may not possess the same moats as other types of business services. The flexibility of the models and their ability to communicate using non-technical language makes it easier for customers to jump from one to another relatively seamlessly. That said, many of the hyperscalers are attempting to pair their AI services with existing cloud services – a famously sticky business. That could prove an enduring advantage.
As with many things in AI, it’s early days and hard to be definitive about what the market for inference might look like in the future. In our view, while some of the dynamics that traditionally favour scale in tech are also present in AI, there are enough key differences that a winner-take-all outcome is not a foregone conclusion. It is perhaps not even the most likely outcome.
While there is the distinct possibility that only a small number of models win this race, our analysis ultimately makes the case that there is a decent chance that the AI space instead becomes a diverse, competitive marketplace. Regulators might prefer such an outcome given concerns about the disruptive power of AI and could take a more proactive approach to M&A than in the past to ensure no single player has outsized influence.
A more competitive marketplace that drives down the cost of inference could help accelerate AI adoption. The MIT study noted above suggested substitution toward cheaper, open models would save the AI industry $25 billion annually. Those savings might come at the cost of lacklustre returns for developers that are investing heavily in closed, frontier models.
Whatever the degree of market concentration, there are sure to be winners and losers, and some malinvestment in models that can’t compete. But that is separate from whether the significant investment that is going into AI infrastructure like data centres and power generation will ultimately be of economic value. While there were many losers from the dot-com crash, the infrastructure that was put in place during that era turned out to be a useful foundation for today’s internet and cloud computing.
-JN
AI unemployment concerns
We have written before about the entirely valid concern that artificial intelligence will hurt the labour market. In September, we observed that recent graduates in the most AI-exposed professions were seemingly faring slightly worse than normal.
In March, we laid out a variety of long-term scenarios for AI, including the possibilities that AI would and wouldn’t enduringly damage the labour market. We actually think it is more likely that profound damage will be avoided, but both are possible.
Optimistically, between 1870 and today, the average hours worked per person in the United States fell by approximately 40%, and yet – significantly thanks to new technologies and productivity gains – incomes rose and the unemployment rate remained low. It is far from impossible that the next great technological leap – presumably, AI – does the same for us. Higher salaries, low unemployment and more leisure time wouldn’t be half bad.
Still, AI is evolving so quickly and there are still so many conceivable scenarios that it makes sense to remain wary and to closely track the subject.
Little trouble in the top-down data
Happily, evidence of mounting job losses proves elusive in the top-down U.S. economic data. Weekly jobless claims have remained not just low but are on a slightly declining trajectory – a remarkable achievement when considered not just in the context of AI but also the energy shock (see next chart).
U.S. jobless claims are low and steady
As of the week ending 04/25/2026. Sources: U.S. Department of Labor, Macrobond, RBC GAM
The Challenger survey of major U.S. job cuts has been choppier than usual in recent years. But it is relatively subdued at the moment despite prominent layoff announcements from such giants as Microsoft and Meta (see next chart).
Similarly, the U.S. government’s WARN Notices – companies are legally obliged to inform the government of impending major layoffs and give 60-days’ notice – is low. If anything, it is trending slightly downward.
U.S. mass layoff announcements and announced job cuts remain lower
As of March 2026. Sources: openICPSR (Inter-university Consortium for Political and Social Research), Challenger, Gray & Christmas, Inc., Macrobond, RBC GAM
Admittedly, when one digs into the Challenger data, it is possible to find individual sectors for which layoffs are clearly rising (see next chart). The technology sector is notable, and there is also a slight uptick in financial and telecom job losses. But by definition some sectors will always be worse than the average and others better.
Layoffs rising in technology, financial and telecom sectors
As of March 2026. Sources: Challenger, Gray & Christmas, Inc., Macrobond, RBC GAM
For their part, the tech giants say their layoffs are to allow them to invest more money into AI development. However, one might imagine that as early adopters of AI, they may also secretly think they can accomplish their goals with fewer human workers. It is hard to distinguish between the two motivations right now.
Gauging employment vulnerability by sector
We are not the first to attempt to identify the vulnerability of workers by sector. Most of these efforts focus upon which sectors are in the best position to harness artificial intelligence to save money and become more efficient.
We think a second, less examined variable is also important.
Jevons Paradox observes that as technological improvements increase the efficiency of resource use, the overall consumption of that resource can actually go up rather than down. In the context of AI, as companies lean more on artificial intelligence, they may need fewer workers.
But if AI succeeds in lowering a company’s overall costs and making them more efficient, some part of that saving will be passed along to their customers. Some sectors have a sufficiently high price elasticity of demand that the lower price they charge could allow their businesses to grow quite significantly. Ultimately, they may require as many or more workers than before.
New technologies have historically created as many new opportunities and even sectors as they have destroyed. That could happen again this time, though admittedly with less certainty than with past waves of technological change.
Some sectors are quite price elastic, including luxury goods, recreation/entertainment, air travel, consumer durables, advertising/marketing and restaurants. If the price drops a bit, demand rises a lot – potentially protecting worker jobs. Conversely, some sectors are quite price inelastic, including cement, medical care, pharmaceuticals and food. Even if AI were to cut costs profoundly, demand wouldn’t rise by much and so workers are more adversely affected.
We combine the AI potential of each sector with its price elasticity of demand in the table below, assessing the scope for job displacement across 31 sectors. The conclusions are not shocking, but nevertheless useful. Broadly, knowledge-intensive white-collar jobs are the most vulnerable, while goods-producing workers are more secure.
Knowledge-intensive white-collar jobs are most vulnerable
As at 05/04/2026. Source: RBC GAM
How to reconcile this analysis with the earlier argument that AI may not permanently increase the unemployment rate? New technologies have historically created as many new opportunities and even sectors as they have destroyed. That could happen again this time, though admittedly with less certainty than with past waves of technological change.
-EL
Canadian housing regulation aims to boost supply
Canadian homebuilding activity has picked up in the past five years, finally exceeding cycle highs from the mid-2000s. This has been led by strong growth in multi-unit construction (see chart below). This acceleration nonetheless failed to keep up with an immigration-driven population surge. A record 482,000 net new households were added in 2024. However, the pace of household formation slowed sharply in 2025 and is expected to remain below-average in the coming years, giving supply a chance to catch up.
Canadian housing starts have picked up in recent years
As of Q4 2025. Sources: Canada Mortgage & Housing Corporation (CMHC), Macrobond, RBC GAM
The increase in multi-unit construction has itself been driven by a sharp rise in purpose-built rentals (see chart below). This has occurred in part as various levels of government have prioritized rental supply in an effort to ease affordability pressure for lower- to middle-income households. A significant correction in the condo market has also seen some developers pivot toward rental units.
New supply and tepid demand pushed the national vacancy rate for purpose-built rentals above its 10-year average in 2025, although average rental prices increased by more than 5% for a fourth straight year.
Rentals now account for more than half of urban housing starts
As of March 2026. Sources: CMHC, Macrobond, RBC GAM
The pickup in purpose-built rentals and what the Canada Mortgage and Housing Corporation (CMHC) calls “missing middle” housing is much needed. This includes ground-oriented housing types typically characterized by gentle- to-medium-density, offering better affordability than single detached homes. But record low single family starts will pressure affordability in a key segment of the market.
Our measure of resale housing affordability has improved somewhat amid falling home prices and lower interest rates. However, it continues to point to poor affordability with prices still 17% above “normal” levels at the end of last year (see chart below). By our math, higher home prices have contributed twice as much to the net deterioration in housing affordability over the past five years as higher interest rates have, pointing to a supply-demand imbalance.
Housing affordability is improving but remains poor
As of Q4 2025. Fixed floor imposes a minimum ‘normal’ mortgage rate in the affordability calculations and so reveals how affordability would look at ‘normal’ mortgage rates. Sources: Canadian Real Estate Association (CREA), Statistics Canada, Haver Analytics, RBC GAM
The Parliamentary Budget Office suggests housing completions need to rise to 290,000 net new units annually to close the housing supply gap by 2035. CMHC has set an even higher bar, suggesting 430,000-480,000 net completions will be needed each year to restore housing affordability over the next decade.
Based on the latter, the federal government has set an ambitious target of doubling the pace of home construction over the next decade, building nearly half a million homes per year. Its main initiative in pursuit of that goal is Build Canada Homes. Announced last fall, this federal agency will support the construction of affordable (particularly non-market) housing. The agency received an initial capitalization of $13 billion and will seek to leverage private capital as well.
When it comes to market housing, there is a disconnect between what home buyers can afford and what developers need to charge to cover construction costs and other fees. Costs have more than doubled over the past decade, led by higher material costs (see chart below), while development charges can add more than $100,000 to the price of a new home in cities like Toronto.
Materials prices are a major driver of rising residential construction costs
As at 05/01/2026. Sources: Statistics Canada, RBC GAM
To bridge the gap, policymakers at the federal and provincial level have announced policies aimed at reducing development costs, cutting red tape and zoning restrictions, and making new homes more affordable for first-time home buyers. Ontario, where housing affordability is particularly stretched, has been the most active. Here are some of the recent initiatives:
The federal government is providing $1.7 billion in funding to provinces and territories to unlock new housing supply through measures like reducing development fees or levies on new home construction.
The federal and Ontario governments will spend $8.8 billion over 10 years to cost-match housing-enabling infrastructure projects. This will allow municipalities to cut development charges by up to 50%. It builds on Ontario’s existing funding for housing infrastructure, which also incentivizes municipalities to cut red tape. Ottawa is hoping to strike similar deals with other provinces.
Last fall, the federal government eliminated the GST on new home purchases for first-time home buyers (up to $1 million, with the tax phased back in on pricier homes). The Ontario government recently confirmed it will do the same for the provincial portion of HST, although only for the next year. So far, other provinces have yet to follow suit.
The Ontario government will allow land to be subdivided into smaller lot sizes to improve housing supply and affordability. But critics suggest the zoning changes don’t go far enough to encourage more medium-density housing. In November, British Columbia expanded its small-scale multi-unit housing rules in a similar effort to boost medium-density construction.
The Alberta government has proposed changes to expedite approvals and speed up housing development, although Calgary recently repealed 2024 re-zoning laws aimed at increasing density.
Demand-side measures like tax cuts risk exacerbating the affordability challenges they aim to address. However, there is clearly an effort to boost supply by funding infrastructure, cutting red tape and reducing development charges. Cooperation between the federal and Ontario governments in this area is encouraging. Similar efforts by other provinces would make us more confident that the country is on a path to closing the housing supply gap and restoring a degree of housing affordability.
-JN
Canada’s new sovereign wealth fund
Having switched to a fall budgeting schedule, the federal government released its first spring economic update in April. It turned out to be more of a fiscal update than a mini budget with relatively few new policy announcements. Positive “economic and fiscal developments” – most significantly stronger income tax revenue assumptions – let the government increase spending while leaving future deficit forecasts little changed. Last year’s budget shortfall even came in $11 billion smaller than projected.
The update included $33 billion in net new spending over the next 5 years alongside another $16 billion in measures announced since the fall budget. That includes a $12 billion groceries and essentials benefit announced earlier this year and $2.4 billion to temporarily suspend the federal fuel excise tax. This tax relief will lower gasoline prices by $0.10/L until Labour Day.
The government’s capital investment plan was little changed. It continues to show a significant pickup in public CapEx this year (an incremental $20 billion on a cash basis).
One initiative that didn’t impact the deficit projections but still drew plenty of attention was the announcement of a sovereign wealth fund called the Canada Strong Fund. Here are some of the initial details:
The fund will receive an initial endowment of $25 billion over 3 years on a cash basis (but considered an equity investment for budgeting purposes). This will grow as returns are reinvested and as the government potentially allocates other assets to the fund.
The fund will operate at arms-length from the government and participate alongside other investors on a commercial basis. It will focus primarily on equity investments in strategic Canadian projects and companies with a mandate to deliver market-rate returns.
The fund will focus on complementing efforts where the government is active through federal agencies and Crown corporations, including the Major Projects Office and Building Canada Act. These aim to accelerate the approval and development of nationally important projects.
Beyond $25 billion in taxpayer money (or money borrowed on behalf of taxpayers) being used to seed the fund, Canadians will be able to directly participate through a new, retail investment product. Details are still being worked out, but the government claims “investors will be able to share in the upside, while their initial invested capital will be protected.”
The Canada Strong Fund differs from many of the world’s largest sovereign wealth funds (see chart below). These typically invest surplus revenues from natural resources (Norway’s GPFP and a number of Middle Eastern funds) or accumulated foreign exchange reserves (several Asian funds). These funds largely invest outside their home countries given limited domestic investment opportunities relative to the size of the funds, efforts to diversify their exposure, or to prevent domestic currency appreciation by holding foreign assets. For example, Norway’s GPFG invests just 2% of its assets in Nordic markets.
The 10 largest sovereign wealth funds each manage more than $500 billion
Accessed May 3, 2026. Sources: Sovereign Wealth Fund Institute, RBC GAM
Much of Canada’s resource revenue accrues to provincial governments. Several provinces have their own wealth funds – most significantly Alberta’s Heritage Savings Trust Fund, which was set up with an initial investment of $1.5 billion in 1974 and now manages more than $30 billion. Despite Alberta’s greater oil production, that’s just a fraction of the size of Norway’s sovereign wealth fund (set up in 1990). The province has prioritized returning resource revenue to taxpayers via low income-taxes and no sales tax.
The federal government collects some royalties from offshore projects and corporate income tax from resource companies. However, it has run budget deficits for the past decade and thus doesn’t have any surplus revenue to invest. It has some $800 billion in financial assets (including taxes receivable, foreign exchange reserves and loans to Crown corporations). It also has more than $100 billion in non-financial assets. Some of the latter, including airports, may potentially be monetized to fund the Canada Strong Fund.
Canada has major institutional investors in the Maple 8 pension funds which manage combined assets worth more than $2 trillion, although they have diversified significantly in recent years and now invest roughly three-quarters of their assets outside the country. The Caisse de dépôt et placement du Québec (CDPQ) has a dual mandate to optimize returns and contribute to Quebec’s economic development. It aims to reach $100 billion in investments in the province this year. But that still falls well short of the Canada Strong Fund’s proposed domestic concentration (the CDPQ’s total assets amount to more than $500 billion).
Investing in projects that the federal government is keen on advancing seems like a winning strategy, with private sector involvement perhaps reducing the risk of cost overruns that tend to plague public projects. But it raises the question of whether these projects can’t be financed entirely by the private sector based on their own merits as determined by the market. While the government says the fund will be operated at arms-length with an independent board of directors, its focus on investing where the government is active suggests some risk of politicization. And guaranteeing return of principal for retail investors could leave taxpayers on the hook for any losses.
Some of the details of the Canada Strong Fund are still being worked out, and while it appears to be an unusual take on a sovereign wealth fund, we don’t think its potential benefits should be dismissed entirely. At the very least, the government putting extra effort and capital into nation-building projects is consistent with a more pro-growth approach to fiscal policy and a bid to ensure Canada can capitalize on its plentiful natural resources.
-JN
