Michael Burry, the investor of “The Big Short” fame, thinks there’s not enough appreciation for depreciation, and this week posted about his qualms about how the likes of Oracle and Meta are handling the accounting on their GPUs. In short, he says they are understating depreciation by extending the useful life of the chips they’re buying, and this artificially boosts earnings.
|
-
The estimated “useful life” of AI servers for the publicly traded hyperscalers is about five to six years. Their useful economic life — how long they’re actually being used to help make money — may be longer or shorter than that.
-
Companies that spend hundreds of billions to invest in data centers don’t count that money as an expense immediately, but rather record the cost over time as the equipment is used. That’s the depreciation to which Burry refers.
-
Tech companies have argued that their chips effectively get a second life by being repurposed from training to inference, which is intended to coincide with when new flagship models are introduced and put toward the rigors of training. Is that in fact the case? We’re going to find out.
|
Company-specific reports from industry bellwethers would suggest older chips are still very much in demand; Nebius, CoreWeave, and Oracle’s businesses all seem to be able to get more of a second life out of those older chips. Still, just because A100s have been able to stand the test of time doesn’t mean future generations of chips will. Recall, for instance, how Nvidia’s Blackwell ramp was delayed because of overheating issues.
|
|
|
The proof, ultimately, will be in the cash flows over time — or a lack thereof — and how the answers to these questions play out. Are consumers and businesses willing to pay for a non-flagship level of AI compute for certain tasks? Early evidence suggests yes. Are chips physically able to hold up to their workloads for a five-plus-year period? Early evidence also hints at yes. Do changes in which tasks GPUs are being asked to perform radically alter the overall ROI on all this spending? It’s too early to tell.
|
|
|
Turn Market Moods Into Market Moves. |
Everyone’s talking. BUZZ is tracking. Sentiment is a distinct factor that moves the market — but can be misleading without a way to measure it. The VanEck Social Sentiment ETF (BUZZ) offers investors just that: a rules-based way to harness crowd conviction. The fund tracks the BUZZ NextGen AI US Sentiment Leaders IndexTM, a methodology that selects and weights 75 large-cap US stocks by sentiment. That’s achieved by AI and natural language processing to scan finance-specific language across online content, then aggregated into investable insight. Because it’s sentiment-driven, the index is rebalanced monthly to capture shifts in discourse. BUZZ naturally picks up on themes as they emerge, from AI adopters to crypto-treasury plays. Discover how BUZZ can add a behavioral lens to your portfolio to reflect the market’s mood. |
|
|
Tesla shareholders voted resoundingly in favor of CEO Elon Musk’s $1 trillion pay package last week, with nearly 70% voting in favor. Their enthusiasm for investment in Musk’s other company was much more muted, however, as 38% voted in favor, 33% voted against, and a notable 17% of shareholders abstained from voting, recent filings show.
The shareholder proposal had argued that xAI’s AI expertise would complement Tesla’s autonomous driving and robotics endeavors, and would give Tesla investors a stake “in a major AI player, potentially yielding significant financial returns while fostering technological advancements that benefit Tesla’s customers and shareholders.” |
- Tesla’s bylaws “generally consider abstention as votes against,” so the measure failed to win approval. Tesla’s board itself abstained from making a recommendation on the xAI proposal to avoid “related party” conflicts, as Musk is the head of both companies.
-
Tesla Board Chair Robyn Denholm recently told The Wall Street Journal that the board “hadn’t previously looked into xAI’s financials or done any of the due diligence required to make an investment.”
-
Musk can’t make his public company, Tesla, invest in his other, private companies. In July, he posted, “It’s not up to me. If it was up to me, Tesla would have invested in xAI long ago.”
-
Morgan Stanley analyst Adam Jones said the vote was a missed opportunity. “We don’t think investors understand just how important xAI is to Tesla and the broader Muskonomy,” he wrote in a note. “Tesla’s relationship with xAI (financially and strategically) is deterministic to the long-term success of Tesla due in part to the natural synergies of data, software, hardware and manufacturing in recursive loops.”
|
To put that last quote plainly: Tesla continues to pitch itself not as a car company, but as the center of our autonomous future, and AI — especially developed in-house by Musk — is key to that plan. |
| |
The utopian future Elon Musk has described on earnings calls is awe-inspiring. And if Optimus robots can “actually eliminate poverty,” as Musk said during the shareholder meeting, that would be pretty amazing. But as things stand, there’s still widespread doubt that Tesla will even roll out full self-driving tech this year. Perhaps with his new pay package, he’ll be more focused on achieving those major milestones and spend more time at Tesla, though that’s not guaranteed.
|
|
|
As we turn in to the final lap of Q3 earnings season, natural gas stocks have emerged as some of the most overlooked sources of revenue growth for investors, according to a screen we recently ran using FactSet data. |
|
|
Level up your trading with index options |
Whether you’re interested in generating income, protecting your retirement portfolio, or expressing your view on market moves, there’s an index options strategy that can help you achieve your trading goals.
|
|
|
|
