Anthropic & Nvidia – The Market Speaks

Nvidia: gigantic cash machine on sale.

• The case for Nvidia remains strong, as despite whatever happens to Jensen, next year’s sales have already been made, giving Nvidia time to work through its current crisis.
• The Good:
  • First, Demand: which remains as strong as ever.
  • With every company that reports Q3 26, the one thing that has not wavered is demand for AI services and equipment.
  • Hence, as long as Nvidia can keep its eye on the ball (see below), calendar 2027 (FY 2028) is going to be a vintage year with 70% growth in revenues to an almost unbelievable $720bn.
  • This is entirely consistent with the capex plans of both the hyperscalers and enterprises, who are all building AI infrastructure as fast as they possibly can.
  • This means that Nvidia is basically already sold out for 2027, giving it superb visibility and the ability to make the forecast that it did at its last set of earnings.
  • Second, gross margin: is going to recover.
  • When it reported FQ2 27, the company said it expected some weakness from higher memory prices in the near quarters.
  • However, due to the price increases it has put through (but have yet to materialise in sales), gross margins will recover to 75% early in fiscal 2027, but the market is not looking that far ahead and remains laser-focused on Q4 26.
  • This is why the company has received no credit for the sales forecast for fiscal 2027.
  • Third, cash machine: which has to be seen to be believed.
  • In fiscal H1 2027, Nvidia generated $74bn in cash flow from operations that was mostly spent on acquisitions, dividends and share buybacks.
  • Furthermore, unlike other companies, the share count is actually falling, meaning that cash is really being given back to investors.
  • In Fiscal 2028, 70% growth at 75% gross margins translates into $366bn in cash from operations, which is a truly staggering figure.
  • Fiscal 2029 is likely to be higher still.
  • Third, valuation: has reached compelling levels even given that the risk of owning Nvidia has just increased materially.
  • The company is trading on a calendar 2027 PER of 9.8x, which, when factoring in 70% growth next fiscal year, means it is being valued as a commodity memory player.
  • Nvidia is by far the market leader with easily the best and a unique product, meaning that there is compelling value to be had at these prices.
• The Bad:
  • First, competition: who now smell blood in the water and are already circling to capitalise on any slip-up that Nvidia makes while it gets its house in order.
  • The current crisis is going to consume management time and morale, increasing the risk of oversight or mistakes which its competitors will be quick to capitalise on.
  • However, the problem they have is that they are also capacity constrained, meaning that they can’t really take share next year even if they could talk the customers into it.
  • Hence, I am not too concerned about competition here, but if Jensen ends up leaving, this will become more of a problem in the long term.
  • Second, data centre slowdown: which could hurt the growth profile in US data centres in calendar 2028.
  • There is a political backlash against data centres in the USA, which I think means that projects that have not broken ground in the USA are going to take longer to do so.
  • I think that the ones that are under construction will get completed, although the owners will have to make more concessions to the local population (not difficult) in order to get their final permits to operate.
  • The good news here is that demand is so strong for AI equipment at the moment that US data centre pushouts in the USA in 2028 are likely to be easily replaced with foreign data centres that are languishing at the back of the queue and clamouring for Nvidia silicon.  
  • This is a great opportunity for Europe, the Middle East, Canada and Mexico to step up their plans and finally reap some of the benefits of the AI boom that shows no sign of abating.
  • Hence, a slowdown may really bite the neo-clouds, who have debt payments to make, but suppliers like Nvidia et al, can probably switch capacity earmarked for the USA elsewhere.
  • Second, marginal buyers: is a large risk for Nvidia, as the shares are already widely owned across institutions, retail and passive investment products.
  • This raises the question of who will come in and buy more shares that will push the price up so that newcomers who buy here can earn a return.
  • However, the last few days have seen a lot of selling of Nvidia, and so I suspect that it is now less widely owned than it was a week ago.
  • This goes some way towards alleviating this concern, which is one of the things that has kept me on the sidelines until now.
• The Ugly:
  • First, single-man risk: if Jensen goes, I think that the shares will fall further as confidence in even the 2027 forecasts will be damaged even further.
  • Nvidia has substantial single-man risk as Jensen is the founder, has taken all of the key strategic decisions and has complete control of the company.
  • This is clear from the flat governance structure where he has around 60 direct reports and no obvious successor.
  • There will be a large crater left in the management of this company if he goes, with no obvious way that the crater can be filled.
  • On balance, I continue to think that Jensen will end up remaining at Nvidia, but obviously now some things will have to change.
  • Second, governance: has been revealed to be wanting.
  • There will now need to be a clear succession plan and better visibility on how the company is run and who the key executives are.
  • For example, none of the senior leaders who are responsible for key business functions are listed on the governance section of the website, which wasn’t a problem until now.
  • Investors will now want to know how this company continues without Jensen and get to know the key executives, as the unique style of this company is going to need someone from the inside to take over one day.
• The net result is that the current crisis came at the worst possible time, which in turn has created a rare opportunity.
• I have added $500bn in debt to Nvidia’s balance sheet for 2028 to deal with the possibility it has to make good on all of the guarantees that it has provided to help the market grow.
• I have also increased the discount rate from 9% to 12% to account for the increased risk that Jensen leaves and the fiscal 2029 outlook and beyond suffer as a result.
• Even then, I can still value Nvidia at $290 per share, which would put it at 16.6x calendar 2027 PER, which is not particularly demanding when compared to the peer group, which it already dominates.  
• If Jensen stays, then I think that the shares could go much higher than that on a 1-2 year view.
• I have taken a half position in Nvidia with a view to deciding on the other half when we know for sure whether Jensen will be going or staying.

Anthropic: early demand suggests risk appetite is as strong as ever.

• Despite the panic and the correction ongoing in the AI trade, Anthropic is pushing ahead with its IPO and has received enough demand to be able to increase the valuation to $1.5tn, which is a comforting sign for the AI trade.
• Anthropic is going through the last stages of the IPO process, and despite the recent sell-off, a change in the political climate and a potential slowdown in the rollout of data centres, there is still plenty of appetite for a fast-growing and risky asset.
• To be fair to Anthropic, the numbers are not as nutty as some would have us believe.
• I think that its targets are roughly $120bn in revenues in 2027 and $200bn in 2028, against which it is committed to compute costs of $70bn in the next twelve months, roughly $500bn by the end of 2030 and a total of $900bn extending into the 2030’s.
• This equates to costs ramping up to something like $130bn a year, but if the company is going to hit $200bn in revenues in 2028, this is clearly not going to be a problem.
• Investors buying into the IPO clearly agree, implying that the wobble of last week was a wobble and not the beginning of a collapse in the AI trade.
• Hence, this is a good sign that growth remains intact and that 2027 is likely to be another very strong year for the suppliers to the AI industry.  
• I have other issues with Anthropic, such as there being no defensible moat in models and the fact that it is not the lowest-cost producer of tokens, contrary to the claims in its roadshow video.
• I think that OpenAI will be in a position to undercut Anthropic’s pricing in 2027 and still be reasonably profitable while Anthropic loses money at the same price.
• This is the main risk as far as I am concerned, and it is a big one.
• Hence, a valuation of 12.5x 2027 price-to-sales looks pretty rich to me, meaning that there is much better value to be had elsewhere in the sector (see above).