TheChek News Flow

Fashion house also plans more acquisitions to further integrate its supply chain

Chanel plans to increase investment in its retail network and real estate by at least 50 per cent this year, as the French design house competes with other luxury groups in a hot market for prime locations.

The company, which is owned by the billionaire Wertheimer family and headquartered in London, also plans to continue to make acquisitions to further integrate its supply chain after a dozen such deals last year, according to its top executives. 

“We are seizing opportunities in real estate which the current environment is offering. So we will be on the offensive,” Chanel’s global chief financial officer Philippe Blondiaux told the Financial Times. 

“We are also expanding our capacity [and] we are accelerating the vertical integration of our supply chain because we believe this is key to controlling our manufacturing and materials.”

Chanel will be competing in a crowded real estate market as top luxury groups spend billions to secure the most exclusive retail locations for their brands.

Gucci owner Kering last month bought a retail block on Milan’s top shopping street for €1.3bn from Blackstone — Europe’s biggest property deal for two years — as demand from luxury groups helps high-end retail real estate defy a wider downturn.

LVMH, the world’s biggest luxury group by sales, spent roughly €2.5bn on real estate investments last year, including for prize assets on Paris’ Champs Elysées. 

Chanel has also recently splashed out on buildings on New York’s Fifth Avenue and Avenue Montaigne in Paris.

The company, made famous by the pioneering designs of its founder Coco Chanel, has been growing rapidly. Sales hit $19.7bn last year, up 16 per cent against 2022 on a like-for-like basis, while operating profits rose 10.9 per cent to $6.4bn. 

The expansion in recent years has come during a luxury boom that has brought record sales and profits for the sector. Chanel has more than doubled both its revenues and headcount in the past decade, according to chief executive Leena Nair. 

“My priority . . . is to protect what we cherish and what differentiates us while continuing to have the drive of a scaled business. We have tripled the number of countries we are in [and] our distribution network has doubled in the last five years,” Nair said. 

As the industry’s growth slows from the giddy highs of recent years, Chanel is emerging as one of the most resilient brands alongside other top tier players such as Hermès and Brunello Cucinelli, which benefit from their high-end positioning and wealthy client base. 

Chanel said sales growth was in the double digits across all its categories from fashion to handbags to beauty. Europe and Asia grew respectively in the high teens and low 20s, bucking industry concerns about Chinese shoppers as Asia’s powerhouse economy slows, but the Americas remained softer at 2.4 per cent growth.

After already increasing its investment in the business by a hefty 83 per cent last year to $1.23bn, Chanel plans to do even more in 2024.

“I don’t think there is a single market, including the US, which we see as saturated,” Blondiaux said. “The US for us is still an under-developed market for luxury if you look at certain [indicators] on wealth.”

In China, Blondiaux believes Chanel is “under-distributed”, with only about 18 boutiques in the world’s second-biggest economy, far fewer than some of its rivals.

“We have plans to continue to invest in China even though Chinese consumers have resumed travelling” abroad to shop, he said, a trend that was slow to pick up following the country’s draconian lockdowns at the end of the pandemic. 

However, gripes about steep price increases have emerged among some of Chanel’s clients. The average price of luxury goods tracked by HSBC has risen 50 per cent since 2019, while the cost of a classic Chanel flap bag has more than doubled to top €10,000. 

Chanel says its price increases reflect higher costs of materials as well as inflation and will maintain its current policies. Pricing contributed 9 per cent to its sales growth in 2023 and 7 per cent came from an increase in volume, Blondiaux said.

OpenAI and its artificial intelligence rivals get all the headlines, but few venture-backed companies have benefited more from the AI boom than CoreWeave, a cloud provider. That's thanks to Nvidia, which allocated some of its precious AI server chips to CoreWeave rather than large clouds like Amazon Web Services, prompting startups and even Microsoft to collectively commit to spending billions of dollars to rent CoreWeave’s servers.

On Friday, CoreWeave said it raised $7.5 billion of debt from private equity giant Blackstone and others to buy more chips and expand its U.S. data centers. CoreWeave co-founder Brannin McBee, who's based in Montana and handles the company’s capital raises, said it hopes to roughly double its data center footprint this year to around 300 megawatts of capacity.

The Takeaway

• CoreWeave says large enterprises now make up most of its revenue

• CoreWeave’s valuation rose to $19 billion in a financing this spring

• The cloud provider could raise even more debt this year

Cloud providers seldom share such data, but CoreWeave’s disclosure implies its facilities could potentially run a few hundred thousand Nvidia H100 graphics processing units, according to someone with knowledge of GPU power requirements. While CoreWeave’s capacity is tiny compared to AWS or Microsoft Azure, the startup’s projected growth could put it within striking distance of Oracle when it comes to GPU cloud sales.

As a younger cloud entrant, McBee said CoreWeave has designed its GPU server clusters from the ground up, which gives it an advantage over AWS and others that have focused on traditional computing tasks such as hosting websites. “It's sort of like walking into Ford and asking them to produce a [Tesla] Model Y,” he said of AWS and others. (AWS didn't comment on the remark but said last month that its cloud sales accelerated in the first quarter, in part due to generative AI computing.)

The debt deal came just weeks after CoreWeave raised more than $1 billion in equity financing at a $19 billion valuation—up from $2 billion in a funding round a few months after ChatGPT launched and spiked demand for GPUs. CoreWeave's stock has been so hot that six months ago, early shareholders and employees sold $642 million worth of shares to new buyers—an almost unheard of sum for such a young business. The company, which previously catered to crypto miners, is based in New Jersey and employs 550 people.

While the jury is still out on how conversational AI startups will make money, cloud providers' path to revenue and profits is well established. In June last year, CoreWeave told investors that it expected to generate more than $500 million in revenue in 2023, up from about $25 million in 2022, and it projected $2.3 billion in 2024 revenue. (It based most of that projection on contractual commitments from customers.) A CoreWeave spokesperson said the company has been “profitable since day one,” without elaborating.

It isn’t clear what CoreWeave is using as collateral for its new loan—a company spokesperson declined to comment about that. GPU cloud providers have been increasingly borrowing against their GPU servers, which an investor previously called a “new asset class.” There’s risk in doing so, of course, as prices for the GPUs might fall over time as supply rises. And CoreWeave may feel pressure to keep up with the immense capital expenditures its larger rivals are planning.

In an interview, McBee said another debt deal could happen this year and talked about how CoreWeave got ahead of power constraints for data centers, why he believes the company is better than AWS at running GPUs, and how demand for its GPU servers has shifted to large enterprises.

This interview has been edited for length and clarity.

The Information: What does this funding announcement signify?

McBee: The scale at which we’re operating, especially relative to our peers—Amazon, Google, Microsoft, Oracle—this [round is] what brings us into that hyperscaler status. I think the differentiation is that we're an AI hyperscaler [in the same category as the world's biggest cloud providers]. That is evidenced not only by that skill of operation, but who our counterparties are. Our investors, our creditors, or our clients, or suppliers. All these pieces are required to position yourself as a hyperscaler. And that's what you're seeing come into place.

What are you going to spend the money on?

Just as you would expect, this is all going to support infrastructure builds. The [$1 billion] Series C round [of equity financing], that's more like continuing to bring world-class engineering talent into the business, supporting research…and also just to bring in tier one technology investors. [The debt round] is to support demand for our product.

What are the biggest bottlenecks around expanding your GPU capacity?

A lot of it is on the power side. It's something that we've fortunately gotten way ahead of. We started securing the capacity we need to execute on our growth plans beginning in Q1 of last year. I have all the data center capacity I need to hit my 2025 [revenue goal]. I would not be able to secure that data center capacity today if I tried to. It's just simply not there.

[Power] is the bottleneck for the broader market right now. And it's one of our most coveted strategic assets. We were able to identify this impending issue in the market and go out and secure the data center capacity that was needed to hit our growth profile.

How many megawatts of data center capacity do you plan to have by the end of 2024, if you expand to your goal of 28 data centers?

Over 300 megawatts.

And next year?

We have the power, we have the data center space secured. It's well in excess of double that amount [of wattage].

I keep hearing that Nvidia's H100 chips perform differently in different clouds. Why do you claim H100s perform better on CoreWeave than, say, Amazon Web Services?

The difference is the engineering solution that delivers that H100 to the client. In other words, it's really the software.

We built our entire software solution around supporting this AI infrastructure, whereas our competition built their software solution around hosting websites and storing data lakes. And it's a very different outcome.

It's sort of like walking into Ford and asking them to produce a [Tesla] Model Y. They can't produce a Model Y because it's a different mechanism to produce those vehicles. So they produce some other electric vehicle that looks like a Model Y, but it is still very different and it has a different performance level. It's not that our competition can't produce something with four wheels on it that's it's electric.

And if you take the analogy a bit further, you couldn't go into Ford and say, “Just turn one of your manufacturing facilities into a Model Y facility,” right? It's the whole supply chain, the whole business. That's the same challenge that our competition faces. They can't just turn one data center into how CoreWeave does it. It's a comprehensive product that we have built from a no-engineering-compromises stance.

Editor’s note: In a recent blog post, Brian Venturo, CoreWeave’s chief strategy officer, said software was CoreWeave’s “secret weapon” to maximize the performance of GPU servers, including by regularly checking their “health” and allowing customers to quickly spin-up servers and automatically increase the number of servers they use. CoreWeave also uses Nvidia’s Infiniband networking cables to stitch together its GPU servers, whereas other cloud providers use their own homegrown networking or are moving in that direction.

Where's demand coming from?

Last year, 90% of our workloads were for training [new AI models], and our pipeline was primarily AI startups and revenue was predominantly AI startups. Today, I'd say 70% is training [models] this year as we move toward inference [running AI models to handle customer queries], and our revenue is dominated by enterprise and our pipeline is dominated by enterprise. I think that's just a function of your typical technology adoption curve profile. Early adopters are going to be startups, and what we're observing is enterprises moving into the space. It's Fortune 500, Fortune 100 companies, it's technology, it's finance. It's increasingly becoming synthetic biology [firms].

Why do you think CoreWeave has been able to raise so much money, relative to its competitors?

The capital is aggregating to us because of our product differentiation. If a GPU was a GPU across the whole GPU startup space, everyone would be getting this volume.

But they're not. And I think that's a little bit of a testament to not only our product, but who our clients are, and our ability to navigate the financial markets. It is not easy to sit down at the table with the types of counterparties we are working with. They're extremely sophisticated. They are world-class investors. Being able to close transactions of this size, it speaks to who we are as a business.

Are you expecting there to be consolidation in the GPU cloud sector?

I wouldn't say that we have any specific plans for M&A consolidation with the GPU cloud space. There's always opportunities within the software side [for acquisitions], I would say.

Editor’s note: Nvidia has made several such software acquisitions this year, including Run:ai. Those acquisitions could boost the efficiency of Nvidia’s own GPU cloud server rental service, which runs in data centers operated by AWS and other large clouds.

How's the relationship with Nvidia going?

They're a phenomenal supplier. They have a great product and market. [Customers] are all asking for Nvidia; no one's asking for anything else. And this is inclusive of like one- to two-year forward-looking plans. We continue to appreciate the relationship we have there and love being able to bring their infrastructure to consumers.

When will you get Nvidia's new chip, the GB200?

There's immense demand for GB200. We would expect to have availability in early 2025.

Landlords stand to gain from higher rents due to growing demand for storage space

Big data is big business. Storing it is too.

Warehouse owners were already tapping into the demand for space created by cloud computing. Artificial intelligence has turbocharged that trend. Landlords are building bigger, more powerful data centres and a European leasing frenzy, reminiscent of that for last-mile logistics during the pandemic, is under way.

Extra demand for space creates a constraint in terms of the supply of power. New grid connections take years so a glut of space in the short term seems unlikely. This creates the strongest fundamental backdrop in any property sector in 25 years, according to Green Street’s Peter Papadakos.

Data centres tend to cluster around core European business locations: Frankfurt, London, Amsterdam, Paris, and Dublin. These sit on major junctions of communications and power cables. Low latency — the time to send data and receive a response — is key for businesses and also matters for deployment of AI applications. 

Limited supply of data centres is holding back take-up and pushing rents up. Underlying demand is essentially unconstrained. Globally 33GW of space was needed last year, thinks Citi, expected to rise to 100GW by 2030. Europe accounted for about 5GW of that, up about a quarter. Growth is expected to slow down to low-teens in the next few years. US-listed Digital Realty recently raised $1.7bn to fund further growth.

Demand comes from the so-called hyperscalers, big US tech groups that typically take space by the building and on a pre-let (pre-construction) basis. Landlords currently have the upper hand: 15-year leases with a break clause are common. Some tenants will accept paying maintenance and investment costs to secure leasing deals. Average rents across Europe are expected to grow at a high single-digit pace over the next five years, at least twice as fast as other sectors.

The cost of building data centres discourages speculative development. At up to $1,200 per sq ft, these cost some six times more to build than traditional logistics sheds. Most of this is due to power requirements, including back-up generators and cooling.

Low latency supply is most desirable. But demand should also rise outside of hubs, in locations that could be used to train large language models, for example, but not host AI services for users. An example may be the acquisition and development plans for the former site of UK battery start-up Britishvolt near Newcastle by Blackstone.

Crunching through terabytes of data will require lots of big data centres. Watch this space.

Delay from the world’s biggest cloud computing provider gives investors jitters over potential ‘air pocket’ in demand for AI chips

Amazon’s cloud computing arm has halted orders of Nvidia’s most advanced “superchip” to wait for a more powerful new model, as investors fret about a dip in demand between the $2.3tn chipmaker’s product cycles.

The Silicon Valley-based chipmaker unveiled a new generation of processors dubbed Blackwell in March, barely a year after its predecessor Hopper began to be shipped to customers. Nvidia’s chief executive Jensen Huang said the new products would be twice as powerful for training large language models, the technology behind OpenAI’s ChatGPT.

Amazon Web Services, the world’s largest cloud computing provider, told the Financial Times that it had “fully transitioned” its previous orders for Nvidia’s Grace Hopper superchip, which was launched in August, and replaced them with its successor Grace Blackwell.

The company said the move “made sense” given “that the window between Grace Hopper and Grace Blackwell was small”.

Nvidia declined to comment ahead of its quarterly earnings report on Wednesday, citing quiet period rules.

Analysts expect the chipmaker to report that sales tripled in its most recent quarter, driven by a frenzy of spending by Big Tech companies on artificial intelligence technology. But some investors have begun to question how long it can maintain its extraordinary growth streak.

While Big Tech companies have pledged to continue to invest tens of billions of dollars into building out data centre infrastructure for AI this year, “there is anxiety [on Wall Street] about a pause in front of Blackwell”, analysts at Morgan Stanley wrote in a note to clients this week.

Production of the new Blackwell chips will ramp up over the course of this year. Analysts expect them to be delivered in the fourth quarter. In the interim, analysts at Citi pointed to a “potential air pocket” in demand for AI chips, after last year’s long wait times for Nvidia’s chips dissipated.

Chips based on Nvidia’s Hopper architecture, such as its sought-after H100 graphics processing units, went into full production in September 2022. The Grace Hopper superchip, also known as GH200, packages several H100 GPUs alongside high-speed memory, connectivity and a central processing unit.

Its successor, the GB200, is the first product to take advantage of Blackwell.

Neither Amazon nor Nvidia would confirm the value of the order. Analysts at HSBC have estimated a GB200 chip, which includes two B100 chips, will cost as much as $70,000, with the price of an entire server carrying the new technology running up to $3mn.

AWS continues to offer other Nvidia chips including H100s to its cloud customers. But, as one of Nvidia’s largest customers, the AWS move is likely to concern investors already worried that tech companies would delay purchases as they await Blackwell’s release.

For much of last year, demand for Nvidia’s H100 chips vastly outstripped supply as the launch of OpenAI’s breakthrough ChatGPT triggered a wave of investment in AI infrastructure by cloud and internet companies, start-ups and corporate buyers.

But since the start of 2024, the long waits for delivery of H100 chips have eased.

Although Nvidia’s stock has almost doubled in value since the beginning of the year on confidence among investors about demand for AI chips, the shares have struggled to show a consistent increase since the chipmaker’s GPU Technology Conference in March where the Blackwell was unveiled.

Nvidia has battled in the past to manage supply and demand smoothly between product upgrades. Recent surges in demand for chips suitable for video games and cryptocurrency mining during the Covid-19 pandemic were followed soon after by a glut of GPUs.

Analysts at Morgan Stanley said they were confident that, even if the supply of Nvidia’s last-generation chips was more readily available now, “in the interim we are seeing new cloud, enterprise and sovereign customers taking all of the available Hopper supply”.

DAX:

MDAX:

SDAX: