Bloomberg Surveillance TV: October 20th, 2026

Angelo Kourkafas · Edward Jones

AI investment cycle has further to run; demand still solid, though financing costs are rising.

He sees no sign of AI demand slowing. What has changed is that financing costs are rising, which pushes out marginal players and moves the theme toward the deepest-pocketed companies, which will have to finance most of the buildout. He argues investors should manage concentration risk while staying with AI, and he expects some benefits to broaden beyond the tech leaders as the theme matures.

“There are no signs yet that demand is slowing. So the demand story is still solid and there. What has changed is financing costs are becoming higher, and that is pushing some of the marginal players.”at 2:50

Edward Mills · Raymond James

A Texas governor win by Hinojosa could bring a data center moratorium, a major risk investors ignore.

Raymond James tells investors to watch the Texas governor's race, not only the Senate race, because it will shape AI and data center policy. He says a win by Gina Hinojosa could bring a real moratorium on data centers. He describes that as a potential category five storm for the data center market that most investors are not paying attention to.

“The gubernatorial race is going to have a huge impact on policy for AI in data centers. If Gina Hinojosa wins, you actually have a real moratorium that could be coming on data centers.”at 12:14

Dave McKay · Royal Bank of Canada

AI model optimisation and in-house compute make long-term data center off-take hard to predict.

RBC runs open-weight models, routes tasks to cheaper models, and has built its own tech stack, compute and storage, so it often does not consume third-party Anthropic tokens. That optimisation means the market struggles to predict what demand will be, while capacity is being built at the same time. He says the key question is what the true long-term off-take is once optimisation is factored in, and this supply and demand uncertainty has created a lot of confusion in the market.

“We're running open weight models, we're running rerouting towards, you know, lower models for this task versus higher cost models. We've built our own tech stack and our own compute and storage capabilities.”at 22:38

Dave McKay · Royal Bank of Canada

AI buildout has drained equity, high-yield and bank funding, squeezing IG capacity and lifting yields.

Demand for AI infrastructure is insatiable, and banks have put large amounts of balance sheet into funding it. The buildout has tapped equity markets, high yield and bank balance sheets, so people are asking how much capacity there is. McKay says this is squeezing out other investment-grade funding and backing up yields. RBC has lent into data centers, is taking some risk off its book through significant risk transfers, and is repositioning its portfolio.

“You've tapped equity markets. You've tapped high yield markets. You've tapped bank balance sheets at the end of the day... You're squeezing out other investment grade funding capabilities. Yields are backing up because of that.”at 24:01

Dave McKay · Royal Bank of Canada

Canada's natural gas power, clear federal rules and telecom links are drawing data center investment and long-term deals.

McKay argues Canada avoids the pushback seen in the US because it has strong capacity to produce electricity, natural gas to power generation, and a clear federal framework. It also has strong telecom networks to move data across borders. He says Meta and others know the cost structure and are signing long-term deals in Canada, and US interest in building AI infrastructure there is growing because approvals are blocked in the US.

“We've got natural gas to power the generation into that. The federal government has created a very clear set of rules that they've syndicated with Canadians to say this is how we're going to build out the data center capacity.”at 24:49

Angelo Kourkafas · Edward Jones

Earnings, not valuations, drive markets; 30% S&P earnings growth unlikely to persist but markets can still hold up.

Strong corporate profits make it hard to be very negative. He expects the roughly 30% S&P earnings growth rate to slow, but argues the market does not need the same growth to persist while economic conditions stay solid. He cites a broadening of earnings and a solid labor market. Valuations have been reflecting macro headwinds, and the tech sector has re-rated lower, so growth is not being extrapolated. He sees this as good news because investors are not paying peak multiples on peak earnings.

“The fact is that companies continue to generate very strong profits... is this 30% earnings growth rate for the S and P gonna be sustained? Probably not. But the market doesn't have to have the same growth rate persist as long as economic conditions remain solid.”at 4:28

Angelo Kourkafas · Edward Jones

Bond yields are reacting to inflation shocks such as oil, and Fed uncertainty adds a premium.

Bonds have been responding to inflationary shocks, and oil prices have been driving benchmark yields, while the growth backdrop has been solid. He also says the Fed's lack of clear guidance adds a premium to bonds. He concludes that the inflation shock is not enough to trigger a major wave of inflation given labor and housing data.

“Bonds have been responding to inflationary shocks while the growth backdrop has been pretty solid.”at 7:14

Angelo Kourkafas · Edward Jones

Growth deceleration, not just inflation, will weigh on markets and bonds in 2027.

Bonds have been responding to inflationary shocks while growth has stayed solid. He argues markets will have to absorb the second derivative of growth, a slowdown that is likely coming in 2027 and that bonds are starting to reflect. This is a mid-cycle adjustment rather than an inflation problem, and he says it should not be a big surprise to markets.

“One of the issues that markets are gonna have to deal with and absorb in 2027 is that second derivative that slowed down in growth. So I think bonds are going to be starting to reflect, and maybe that has already started, some of that deceleration.”at 7:26

Angelo Kourkafas · Edward Jones

Base case: Fed hikes two more times, reversing 2025 cuts; a mid-cycle adjustment, not a 2022 scenario.

His base case has the Fed hiking twice more from here, removing the 2025 cuts while keeping the 2024 cuts. He argues this is not like 2022, when the policy rate sat far below inflation, and that the current policy rate is not in that position. He sees the Fed's uncertainty as adding a premium to bonds, and says a couple of hikes can restore confidence that policy is calibrated to an AI-supported economy. He says labor and housing data do not point to a major inflation wave.

“Our base case calls for the Fed to basically hike two more times from here, removing the rate cuts that they delivered in 2025, but keep the 2024 cuts in place.”at 7:55

Edward Mills · Raymond James

Democrats likely win the House and possibly the Senate, needing to flip two or three Trump-won seats.

Raymond James's analysis says Democrats likely need to flip two or three seats that Trump won in 2024 by 10% or more. He notes such margins have been rare: the last time a state went that far was 2018, and only three of the last five elections had it. Texas, Alaska and Ohio are the key cases, and the polling suggests Texas could go blue, though he says that usually requires a weak candidate. Midterm polling has been closer than presidential polling, so if the polls are right, Democrats do well.

“Democrats likely need to flip two or three seats that Donald Trump won in 2024 by 10% or more.”at 11:06

Edward Mills · Raymond James

AI regulation is in its first inning; Democrats likely to prioritise AI and data-center rules quickly.

Democrats are likely to prioritise AI and data center regulation quickly, possibly in the lame duck session or right after taking control in January. He argues that AI and data center opposition has become a winning electoral issue, which pushes members of both parties to adopt it. Changes could cover environmental rules, power access and how models are regulated, and he says the country is only in the first inning of this.

“...changes to the way in which we regulate different models. I think we're in the very early innings. We're probably in the first inning of all of this AI regulation.”at 15:39

Edward Mills · Raymond James

A Democratic sweep could push Republicans to fund defense via a lame-duck reconciliation bill.

Raymond James wrote that the worse Republicans do on November 3, the more likely they are to use the lame duck session for a reconciliation bill to increase defense funding. A supplemental request has cleared its first House hurdle, and there is a push to raise it toward what Trump has proposed. The reconciliation bill that funded the defense budget last year covered special projects rather than regular defense needs, so a third reconciliation bill would be the chance to let non-primes compete. He calls it Republicans' last chance to lock in defense spending for the rest of Trump's term.

“The worse Republicans do on November 3, the more likely Republicans try to use this lame duck session to do a reconciliation to increase defense funding.”at 16:15

Dave McKay · Royal Bank of Canada

US–Canada trade friction is hurting confidence, home buying and investment; Canada is diversifying exports.

McKay says the US–Canada relationship is in an adjustment period, with trade friction hitting steel, autos and softwood lumber. Canadian consumers are not buying homes because of confidence, not price. Investment is slowing, and foreign investors and manufacturers are waiting to learn the rules of the game. Canada is diversifying with new energy and LNG links to Asia and Europe, and he expects it to find a way to keep trading with the US.

“It certainly affected us in that Canadian consumers aren't buying homes. It's not a price issue anymore, it's really around a confidence issue.”at 20:56

Dave McKay · Royal Bank of Canada

AI-driven cyber attacks are mounting in complexity, and RBC's defence costs have risen substantially.

Firms have to play offense by rebuilding their businesses with AI and play defense against cyber attacks, which are growing in number and complexity as the models get more sophisticated. RBC has run its internet-facing and market-facing software through the most advanced models available, including Anthropic's, and reinforced it. That has cost a fair bit of money, and the cost of RBC's defense has gone up substantially. He calls it an evolving game that requires continuous work.

“Playing defense against cyber attacks, which are mounting the complexity of those cyber attacks and the complexity of those models. So in our RBC playing defense, certainly, the cost organization has gone up substantially.”at 27:43