Heavy hitter: Anthony Scilipoti’s next target? AI

Anthony Scilipoti plants his feet in a staggered stance, raises black boxing gloves to his sweat-streaked face and looks me in the eye. “What do we do next, coach?”

We’ve been at it for 40 minutes in Scilipoti’s backyard north of Toronto, since before the sun broke above the horizon—a version of his weekly training routine modified to account for me, a pugilist newbie. “Coach” is normally Mario Lechowski, a former lightweight boxing champ with an eastern European accent and a penchant for profanity. When Lechowski attempted to give me my first F-bomb-laden boxing lesson earlier, his exasperation poured over with a foreign phrase that the internet later told me was a rather versatile Polish curse word. Nonetheless, I’ve donned Lechowski’s boxing pads, turning me into Scilipoti’s human punching bag.

At 5-foot-11 and 225 pounds, Scilipoti, a forensic accountant and CEO of Veritas Investment Research, Bay Street’s largest independent research shop, is surprisingly fleet of foot, having trained with Lechowski for five years. Each session starts with several minutes of skipping and ends with 55 push-ups—one for each year of his age. In the winter, the action moves inside, to Scilipoti’s personal gym, where he can bench press his own weight.

All of which is to say that as Scilipoti begins wailing on the pads I’m holding unnervingly close to my face, a thought crosses my mind: So this is what it feels like to take a punch—literal or figurative—from Anthony Scilipoti.

Under his guidance, Veritas has made a name for itself by making contrarian calls against some of corporate Canada’s most cherished stocks, slapping them with “sell” recommendations when other Bay Street analysts see only upside. More than a quarter-century ago, when Veritas launched, that meant warning investors away from the hottest and largest company in Canada, Nortel Networks, which then famously accounted for more than one-third of the Toronto Stock Exchange’s benchmark index. And over the years, Veritas has delved deep into the financial statements of many other corporate giants, including Royal Group Technologies, Valeant Pharmaceuticals (now Bausch Health), SNC-Lavalin (now AtkinsRéalis) and Bombardier, to name a few, divining troubling patterns and gaps in the numbers.

Jab. Right cross. Left hook to the body.

More recently, Veritas has been taking aim at what might be history’s hottest trade: the artificial intelligence revolution, and specifically the hundreds of billions of dollars in annual spending that underpins deals between chip makers, data centre builders and AI service providers. Many of the same characteristics that marked the Nortel era—the overhyped exuberance and money flowing in circles between suppliers and their customers—are echoed in today’s headlines. “It’s like a party that’s getting long in the tooth,” says Scilipoti. “Nobody wants to say it’s going to end, and everybody wants to keep it going, but we’re almost running out of alcohol.”

Yet while its warnings about AI date back to 2023, Veritas has yet to recommend selling high-flyers like Nvidia and Broadcom. That restraint was hard-won, Veritas having been early on several of its past calls. With Valeant, it reiterated its sell recommendation roughly 10 times, even as the stock doubled, before reality, and the stock price, came crashing down, wiping out 90% of Valeant’s market value. “Things can make no economic sense for a long time,” says Scilipoti, “until suddenly they matter.”

After the training session has blessedly ended, and the muscles in my arms have assumed a Jell-O state, I ask Scilipoti what draws him to boxing. “It’s not just physical. It’s mental,” he says, as Euro Cash Flow—Euro for short—the small white Coton de Tuléar he shares with his wife, Ana, and their three daughters, scampers around our feet. “No matter what you’re doing, you’re always in a fight with yourself, and the better you can control yourself, the better decisions you can make.”

What’s his battle with himself?

“I have to battle my tendency to be bearish, to always look for problems, because I’m always wondering what could go wrong,” Scilipoti says. “And overweighting that negativity can lead me to a blind spot and not noticing the positives.”

After 25 years of finding problems, his trickiest call may be deciding when the market will finally care about the ones he’s found under the AI hood.

For now, he’s keeping his gloves up.

Step through the doorway into Scilipoti’s modest office on the 31st floor of a tower downtown (which has a chin-up bar wedged between the frame—he can do eight), the first thing you notice is all the monkeys. Statues of the see-no-evil-speak-no-evil-hear-no-evil primates sit on nearly every surface and stare out from pictures on the walls. The inspiration for Scilipoti’s collection hangs nearby: a framed illustration of John Roth, Nortel’s disgraced former CEO, drawn three times over as each of the so-called Wise Monkeys, a piece Scilipoti bought from the artist after it was published in a magazine in the early 2000s.

It’s a fitting representation of the corporate obfuscation and accounting games Scilipoti and his team have dedicated themselves to exposing.

The type of sleuthing carried out by Veritas’s 12 analysts, several of whom have accounting backgrounds, is often called forensic accounting—think of them as CSIs, but with suits and calculators. Dimitry Khmelnitsky, a long-time Veritas employee, prefers the term “investigative accounting,” and he readily admits that as a business admin student at York University’s Schulich School of Business in 1999, he found accounting rules mind-numbingly dull—until he took a class Scilipoti was teaching that exposed the many ways companies bend those rules. “In theory, one plus one should never equal three,” says Khmelnitsky. But companies “take something seemingly rule-based and black and white, and turn it into an art form, where it can become anything they want it to be.”

Veritas’s analysts spend most of their time sifting through financial statements from top to bottom—paying special attention to the fine-print notes, where the good stuff often hides. That might mean realizing that an earnings figure trumpeted on page 10 excludes a hefty chunk of losses that are explained only in a reconciliation 25 pages later.

Take Valeant. In 2011, the Laval, Que.-based drugmaker was one of the hottest stocks in Canada, with a voracious appetite for acquisitions that garnered universal praise from Bay Street analysts. But when Khmelnitsky delved into the details of those acquisitions, he was astonished to realize that Valeant’s organic growth—sales from drugs it already owned—was half what the company claimed, prompting him to issue an alert that concluded, “Buyers beware.”

The hunt for the truth can take Veritas to some unusual places. To investigate the soundness of Yellow Pages’ revenue growth, Scilipoti had staffers count every full-, half- and quarter-page ad in directories from nine cities over three years, then phone the Yellow Pages call centre posing as would-be advertisers to gather intel on how much it was actually charging for ads. The verdict: Its underlying business was deteriorating but masked by annual price increases.

Veritas used to call these inconsistencies “red flags.” But Scilipoti has since decided that term is too absolute. “A red flag is like a stop sign that means everything is bad,” he says, while sitting on a balance ball he uses in place of a chair.

Instead, Veritas seeks to identify what it calls flammable items—issues that may be harmless on their own but could blow up given the right spark. Some are accounting choices, like booking spending as an asset to be written off gradually, rather than as an immediate cost, which flatters a company’s earnings picture. In the Big Tech world, this means stretching the assumed lifespan of AI chips and servers, which Veritas estimates lifted profits at Meta, Alphabet, Amazon and Microsoft by about 13% in 2025.

Or it could be a custom profit measure that can flatter results. This year, Brookfield Corp. changed how it calculates one such measure—funds from operations—for some of its investments. Veritas says the change retroactively boosted that figure by 13% with no change in cash received.

Other flammable items might relate to the business itself: cash flow that never quite matches reported earnings, a CEO’s mysterious departure, money shuffled between related companies. At Telus, it was mounting debt: By Veritas’s count, it was paying out roughly 1.5 times free cash flow in dividends, while its net-debt-to-EBITDA ratio was almost four, above the telecom’s target range.

The more flammable items pile up, the less it takes to set them off. The spark itself usually comes from outside the company—an economic downturn, perhaps, or a new competitor. Most often, says Scilipoti, it’s the moment investors stop handing over money. He watched that happen with Nortel. By October 2000, Veritas was warning that market thirst for telecom financings was drying up, leaving Nortel’s customers struggling to pay for more equipment. Within months, they began to collapse, taking Nortel with them.

“Somebody wakes up and says, ‘Wait a minute, you’re not generating any cash, but you want more of my cash?’” When that happens, things tend to unravel quickly, he says. “The killer to all stories is liquidity drying up.”

Scilipoti always had a knack for reading people. As a teenager, he started a business called Teeza, selling women’s jeans at the Pickering Flea Market. He noticed that shoppers lit up when they squeezed into a smaller size. So he and his mother, Margaret, relabelled his stock: The 28s became 26s and so on. “It was like magic,” he says of watching jeans fly off the racks. The lesson stuck: “People believe what they want to see.”

Both his parents are Italian immigrants and entrepreneurs: His father, Sebastian, operates a furniture company that built the cherry table in Veritas’s boardroom; his mom still runs an iteration of Teeza. Scilipoti has always believed his business background gave him an edge over other equity analysts who only ever looked at companies from the outside. If a company promises 5% growth, a business owner knows the only ways to do that are to raise prices, sell more product or buy another company. “There’s no magic,” he says. “If a company says it’s growing, you have to test that. Was it accounting ingenuity, or was it business ingenuity?”

But Scilipoti’s inherent skepticism was also shaped by what was then a little-known gold miner called Bre-X. As a young accountant at Arthur Andersen in 1994, he followed a tip from a stockbroker and sunk $5,000 into the speculative play, which claimed to own the world’s largest gold deposit, in Indonesia.

Over the next three years, he watched his stake grow 50-fold, only to see it crash when it was revealed the company was a fraud built on salted core samples. The experience left its mark as Scilipoti watched thousands of investors lose everything. “I can’t stand it when the little guy gets screwed,” he says. But it was a worthwhile lesson. “The worst thing that can happen is you make money on your first investment, because you start to think you know what you’re doing,” he says. “The more you lose, the more you realize you don’t, and it forces you to learn and get better.”

By 1999, Scilipoti was teaching accounting at York and working with forensic accounting legend Al Rosen and his son, Mark, writing critiques of stocks for Mike Palmer, head of research at the brokerage First Associates Investments. The Rosens, Scilipoti, Palmer and Bill Marshall, a veteran Bay Street salesman, joined forces, adopted the name Veritas and, in February 2000, took aim at Nortel. “There is only so much that can be hidden for so long,” the team wrote in that first report, highlighting how Nortel appeared to be financing its own customers’ purchases while flattering its operating cash flow (a pattern Veritas now sees again in the AI boom).

They kept up the criticism even as Nortel’s shares bounced higher. Being the only research shop calling it a sell drew heckles from brokers and money managers—the exact people Veritas hoped to eventually sell its research to. “You wouldn’t believe some of the calls I got, screaming, ‘We’ll never, ever deal with you guys,’” recalls Marshall, who signed on to market Veritas’s reports.

When the stock collapsed spectacularly starting in July 2000, however, clients realized there might be value in listening to contrarian cranks. The following year, Marshall (who retired in 2007) secured an arrangement with RBC Dominion Securities that gave its wealth advisers exclusive access to Veritas research, locking out the other bank-owned brokerages. Pension funds and other institutional money managers came on board, too.

In 2001, the father-and-son Rosens left Veritas to start their own firm, Accountability Research Corp., after the two sides had a disagreement about the direction of the business—a rift neither Scilipoti nor the Rosens want to talk about, even 25 years later. The normally outspoken Scilipoti will only say the two sides agreed to part ways. “I have absolutely no hard feelings, and I wish them well,” he says.

The RBC wealth management relationship continues to this day, with Veritas also running training sessions for rookie advisers. During a celebration for Veritas’s 26th birthday, Scilipoti thanked the firm’s clients “for standing by us over these 26 years…especially Royal Bank. God bless the Royal Bank, and we also want to thank the Royal Bank.”

For its part, RBC Dominion Securities senior VP and managing director Mike Scott says that the bank “has always been proud of the depth and quality of research we provide, and Veritas is a key component of that.” It’s also been instrumental in helping new advisers properly understand financial statements, he adds—“an absolute core capability for any adviser.”

Still, launching Veritas as an employee-owned, independent firm flew in the face of how analyst research was packaged and sold—it was mostly a sweetener brokerages threw in to win trading and investment banking business. The idea that money managers would pay for research was a hard sell.

That pitch has only gotten harder. When Mike Binette started as an analyst at Burns Fry in 1983, it “was a noble profession,” says Binette, now a career coach in capital markets. Finding six-month-old production figures in a Japanese trade magazine made you a hero, he says. “Now, you can probably get what the production was at the plant last night.” Data has been commoditized, and so has the job of being an analyst. “It’s really difficult as a pure analyst to say, ‘My material is so unique and different that someone is going to pay me for it.’”

As for Veritas, Binette adds, “I frankly don’t know how they survive.”

Veritas’s business has been squeezed in other ways, too. Institutional clients have traditionally paid for research out of the commissions they generate on trades, and that pool has shrunk thanks to falling commission rates, the shift to ETFs and mergers among money managers.

So Veritas has branched out. An asset management arm, launched in 2016, runs three funds of its own, with combined assets of about $200 million. And a foundation arm applies the forensic approach to charities, rating them for donors on governance, impact, strategy and financial stability. But research is still the core, and Scilipoti says it has to earn its keep every day. “Tomorrow at 9:30, the stock market’s going to open with or without Veritas,” he says. “If we’re not adding value, we will not exist.”

In the early days, almost every Veritas report took a dour view, from IMAX, TLC The Laser Centre and Royal Group to Geac Computers and Call-Net. And it’s worth noting that other than IMAX, all those businesses have disappeared or were taken over at lower valuations.

When Scilipoti penned a negative report on a tech player in the early 2000s, the company reported Veritas to regulators and accused it of shorting its shares. The regulator investigated Scilipoti’s trades and positions. “I didn’t personally short a single stock, and I didn’t own a single stock,” he says. “I’ve never traded in names I write about.”

Still, angry letters and lawyer threats poured in, now framed and hanging proudly on Veritas’s walls.

The first came from Bombardier in 2001. The company’s lawyers had found real mistakes in its reports—a misspelled word here, a transposed digit there. Ever since, the firm’s tolerance for error has been zero. “You could be totally right in your argument,” says Scilipoti, “but if there’s some error in a calculation, they will hammer you and discredit you.”

The threats can also be costly, both financially and mentally. In 2015, Catalyst Capital Group sued Veritas for $50 million over a report that said its affiliate, Callidus Capital, could only sustain itself as a public company with Catalyst’s support. The suit, by Catalyst financier Newton Glassman, alleged Veritas conspired with hedge fund West Face Capital to defame Callidus, a claim both Veritas and West Face denied. By 2021, the case had lain dormant for five years, and nothing has happened since then, either. (Catalyst didn’t respond to a request for comment.) “We’ve never lost a case,” Scilipoti says, “but you still have to pay the lawyers.”

Such episodes are a strain. “You can feel you did everything right, but when you go to court, it’s what the judge thought,” he says. “It’s extremely stressful. You can’t sleep.”

Khmelnitsky, who served three years in the Israeli army in the 1990s, puts it this way: “I served in a combat unit, and I was way less worried then than I am now when I’m publishing a report.”

Scilipoti hasn’t just criticized companies for touting made-up metrics that fall outside generally accepted accounting principles (GAAP). He also saw a role for Veritas to tighten accounting standards from within the industry.

In 2003, he began volunteering on the boards of Canadian and international accounting standards organizations, and pushed heavily for reforms to clamp down on non-GAAP measures—in one 2016 study, Veritas found 70% of companies in the S&P/TSX 60 index used them in ways that seemed to flout regulators’ guidance. “Anthony put a laser focus on this,” says Linda Mezon-Hutter, a former chief accountant at RBC who’s now vice-chair of the International Accounting Standards Board. “He really got some publicity around it and got all of us to make sure we were doing our part.”

In 2021, Canadian securities regulators enacted binding rules that require companies using non-GAAP measures to also give at least equal prominence to standard accounting figures and show exactly how they are calculated. For the first time in 2027, a new international standard will pull some of companies’ custom profit measures into their audited financial statements.

This is all coming far too late for Valeant investors, who watched $110 billion in market value disappear. One issue, as Khmelnitsky saw it, was Valeant’s aggressive use of non-GAAP numbers. In his July 2014 sell recommendation, he told investors that owning the stock was “a matter of faith” because management didn’t disclose enough information to independently verify the success of its acquisitions. It took 19 months for another analyst to slap a sell on Valeant, by which time the stock was in freefall.

Yet Scilipoti says it was a humbling experience. Flammable items notwithstanding, with rates falling and money practically free, “stupid business models could work,” and clients potentially missed out on gains by selling too early. “Being right and making money with being right,” he says, “are two separate and distinct skills.”

For Veritas analyst Liam Gallagher, Telus’s token dividend increase in November 2025—the latest in a two-decade streak of hikes—was the last straw. His analysis using Veritas’s metrics showed the telecom had generated $1.6 billion in free cash flow over the previous year yet paid out $2.3 billion to shareholders, roughly a third of it in shares. “That’s a huge problem, because it’s 2% to 3% dilutive,” says Gallagher. “You can’t have that in a telecom business that’s growing zero to 2%. You’re not going to create value.”

He became the first analyst to slap a sell recommendation on Telus’s stock. Then, in July—the same month former CIBC CEO Victor Dodig took over from Darren Entwistle—Telus slashed its dividend by 55%, and by early October the stock was down 46% since Gallagher’s call. In April 2024, Gallagher’s predecessor on the telecom file, Desmond Lau, had been the first analyst to call BCE a sell, too, about a year ahead of its own dividend cut in May 2025.

The thing about being first on a sell call that proves correct is that it gets crowded quick. Telus now has three sell recommendations and 11 holds, the latter being about as negative as most equity analysts ever get. Which means Scilipoti has started to ask a question that comes less naturally to a guy like him: What could go right? “When everybody’s on one side of the boat, it’s going to tip over,” he says. “And it’s equally true when everybody has a sell.”

A sell rests on evidence in the financials. A turnaround is built on what-ifs and is much harder to call. Veritas isn’t there yet on Telus. Until Dodig lays out a credible plan to pay down its $26 billion in net debt, says Gallagher, “they’re going to be in the penalty box.” But Veritas is watching for the turn, wary of a mistake it’s made before.

A decade ago, when U.S. hedge funds piled into bets against Canadian banks, Veritas was right there with them. Credit losses were rising, and Scilipoti says parallels to the U.S. housing crash were “alluring.” But the banks never buckled. “Canada is not like the U.S., and Canadians are not like Americans,” he says. Today, the firm is more bullish on the banks than most of Bay Street, expecting fees from blockbuster IPOs and their booming wealth management divisions to outweigh the housing hangover, for now.

That said, Scilipoti is watching housing closely. Each year, Veritas staff pose as buyers at new-condo sales offices in Toronto and Vancouver. This year in Vancouver, Scilipoti himself was offered a discount of roughly $700 a square foot on one unit, provided he sign a non-disclosure agreement. “They don’t want the discount to lower the price on other units in the building,” he says. “It tells you about the state of the economy. Real estate is a big employer. It feeds into other aspects of business.”

It’s also part of Veritas’s most contrarian active call: Brookfield Corp., the real estate and asset management giant. It is, Khmelnitsky admits, “the lonely view on the Street.” Since slapping a sell on Brookfield in 2022, along with most of the companies in its tangled orbit, the stock has marched upward.

That rating, Scilipoti says, is “based predominantly on our findings of circular funding flows, concentrated illiquid exposures and the risk that financial stress spreads between Brookfield entities because of their extensive related-party dealings.” Its real estate disclosures, for instance, indicate that sales to Brookfield insurance helped fund a gap between distributions and operating cash flow. “In our view,” he says, “this reliance on parent funding and affiliated asset sales raises concerns about the quality and repeatability of the real estate contribution to earnings.” (Brookfield did not respond to a request for comment.)

While Khmelnitsky acknowledges the companies’ stocks have performed well, he’d need to see a narrowing gap between what management says and what the numbers show before he’d reverse his recommendations.

Until then, Khmelnitsky knows there’s a risk of missing out on gains. But when the numbers and the market disagree, he bets on the numbers. “There’s a lot of soul-searching involved—what am I missing?” he says. “But you’ve got to go through it, so long as the issues that are of concern continue.”

It’s hard to grasp the scale of the AI buildout now under way given the vast sums being deployed—according to one study out of Columbia University, AI companies will need to collect roughly US$3.5 trillion a year from businesses and consumers by 2032 in order to justify the sums being sunk into data centres alone. That’s close to 9% of the entire U.S. GDP.

Even with so many deals being announced, one in particular caught Scilipoti’s eye. In 2023, chipmaker Nvidia had put about US$100 million into CoreWeave, a little-known AI cloud company that was one of Nvidia’s customers. The stake was relatively tiny, but it pointed to a pattern Veritas had seen before. When a company invests in a customer, the money goes out one door of its cash-flow statement, as an investment. If the customer then uses that money to buy the company’s products, it comes back in another door, as sales and operating cash flow—the figure investors prize most.

That type of equity-driven vendor support has since grown from curious to conspicuous. Veritas’s AI analyst, Benjamin Butler, has counted roughly 80 deals from 2023 to the end of 2025, worth US$49 billion, in which Nvidia or its co-investors bought into companies that are also customers. Then came the big one: a plan to invest up to US$100 billion in OpenAI as it buys Nvidia chips. “The hairs on the back of my neck stand up as an accountant,” says Butler.

Lately, the support has grown more creative. In its most recent quarter, Nvidia gave some customers up to a year to pay, and the amount it’s owed jumped by US$22.3 billion. It agreed to buy up to US$36 billion worth of computing time from cloud companies that are buying its chips. Weeks later, Nvidia promised to guarantee up to US$105 billion for an OpenAI data centre complex in Ohio, on the condition that the site be filled exclusively with Nvidia gear. In Veritas’s view, the company is increasingly the one making sure its customers can afford to buy its products.

Meanwhile, the demand for capital keeps growing. In August, Nvidia tapped six financial giants, including Brookfield, Blackstone and BlackRock, to raise US$500 billion to fund new AI infrastructure. Three of those firms, Brookfield among them, own insurance companies, and Veritas expects policyholders’ money to become a significant source of funding for data centres.

Veritas is careful to say its critique of AI’s circular money flow is “economic, not accusatory.” Nvidia’s auditors have signed off on these deals and, unlike Nortel, it’s fabulously profitable—US$60 billion in its latest quarter alone. But Butler’s numbers show its cash generation weakening just as the support it provides broadens. Last quarter, Nvidia raised nearly US$25 billion in debt, its first big bond sale. Scilipoti has a metaphor for this phase of a boom, when the party’s ending, and alcohol is the liquidity keeping it going: “There’s nothing left, everybody leaves, and you get stuck with the empty bottles.”

Still, Veritas hasn’t put a sell on any AI stock, a reflection of the firm’s evolution. The accounting may support one, but a sell recommendation is really a bet that private markets will stop funding the buildout, and timing that is treacherous. What Veritas can do, says Scilipoti, is flag the flammable items and leave the spark to the market. “There’s going to come a time where we’ll say sell it,” he says. “But for now, we’re just watching the madness.”