Hello and welcome to the September market update webinar. My name is Dan White, one of ARC's portfolio managers, and I will be moderating today's call. This call was pre-recorded on September 11th, 2026. And with that, I will now turn it over to ARC CEO and CIO, Kathy Wood. Kathy.
Okay. Thank you, Dan. Well, in a few weeks, I'll be putting out a letter putting into perspective the environment we believe we're in right now. We do believe that the right comparison is the industrial revolution. And if you look back then there are a number of observations. Real growth accelerates. It accelerated five-fold from where it had been over the previous 400 years. And this is based on research that our very own Brett Winton, our chief futurist, has done in collaboration with academic research. So real growth accelerated five-fold from 0.6 to 3%.
Inflation, we were under the gold standard then. So there was no inflation but the tendency was toward deflation because these technologies telephone, electricity, internal combustion engine were deflationary in nature as all technologies are. And the yield curve was inverted more than 60% of the time on average by 100 basis points. And so you'll notice today that the yield curve is flattening and from a trend point of view it has been flattening since 0809 as we have been evolving towards this new world. It certainly is a new world for those who think that the yield curve inverting and we do believe it will invert is a harbinger of recession. We think quite the opposite. We think we're moving thanks to the five technology platforms around which we've centered our research and investing. So AI, robotics, energy storage, blockchain technology, and multiomics technology in the life science space. We do believe we're in a deflationary world with accelerating growth.
Interest rates are going to be the big change in terms of the direction. So for the 40 years through 21 we have been in a declining rate environment and at the beginning of that decline it was because we were moving from double-digit inflation into much lower than expected inflation. I remember I was there in those early days when Milton Friedman said inflation was embedded in the system at a double-digit rate. Get used to it. Of course that did not happen. Then of course it got down to very low levels because of COVID. The COVID really crash got us down to nearly 0% interest rates. That's not healthy. That 0% basically said the markets weren't working anymore. And we had been in that sort of environment for a while until the Fed jacked up rates 22-fold and really pushed inflation down but also exacerbated the supply chain issues that COVID did cause.
So now we believe that interest rates have turned secularly. They are now we're now breaking into higher highs from a trend point of view and the equity market is taking it better I think than most people would expect because most of the increase in interest rates if you look at long yields and TIPS has been because of real growth expectations not so much inflation accelerating although the recent kick in oil prices has given the inflation expectations a little bit of a lift here, but by far the real growth has dominated this. So we think the market will take a while to adjust meaning two steps forward, one step back which we think is a good thing. The market is climbing a wall of worry. And we point back to the 80s and 90s in this letter as the golden age of equities. Sure, you had corrections and so forth. The market had corrections. But by and large, it was straight up for the market in a trend sense. And we think we're in a similar kind of environment now.
And I know there are a lot of worries. In fact, there are the same worries as there were in the 80s and 90s. Inflation, interest rates, wars, technology shifts, tariffs. Of course, the technology shifts back then were PC and wireless. My goodness, how were those going to change the world? And that they did. So, we're excited about this new world. And we think that our strategies are on the right side of change and we encourage not only investors but individuals maybe students studying teachers teaching people who feel that technology is disrupting or could disrupt their own career. Get on the right side of change because we think the wealth generation opportunities on the right side of change are going to be enormous. So with that, Dan, I'll hand it back to you.
Great. Thank you, Kathy. We will now begin the question and answer portion of today's webinar. Questions were received from our clients, our distribution partners, and sent to our funds X and Instagram accounts at ARC_Funds. The first one, Kathy, sticking to kind of some of the macro topics here. If we have another yield curve inversion, how may investors distinguish a deflationary boom inversion relative to a recessionary inversion in real time?
Okay. Well, this is a good question. And there's going to be a lot of evidence both ways. We think dominating the picture will be very strong real GDP growth. We see what's happening in AI. We've used this example before, but it is shocking that Anthropic's revenue annualized run rate of revenues went from 9 billion in December last year to 65 billion in July. That's we've never seen that before. And I think we're going to see a lot more of those sorts of surprises. We already have. We're seeing the Amazons the hyperscalers their revenue growth reaccelerating and these are very big companies so they are moving the needle and we think that's going to become very broad-based for companies on the right side of change and so we think the deflationary boom will characterize the macro environment broadly defined however there are going to be deflationary busts as well as companies that are not on the right side of change that are being disintermediated, disrupted whether it's in the transportation space or in the telecommunications broadband space. Almost every space in the old world order is going to be disrupted because these technologies are so broad-based and will sweep across sectors and industries.
And so I think there will be debates around oh my goodness is this a bust or a boom but by and large I think the GDP statistics even though today they're based on the industrial age they most of these statistics came out of the industrial age I think the revisions are showing us as they move up that this new age is going to be characterized by much stronger than expected growth by and large. I will say one other thing we're focused on is the private credit and private equity space which has really funded itself or funded its companies with short-term debt with and that is floating of course they could be in a world of hurt because what have they focused on? They focused on more mature companies in the traditional world order with decent cash flows. SaaS a good example and did not expect those to be disrupted. Now put on top of that an interest rate increase. Well maybe that's one of the weakest links in the global economy. And so I think there will be a lot of drama around how rapidly these changes are occurring and who is in harm's way and who is on the right side of change.
Great. Thank you Kathy. Lot to digest there. We'll move over to the more tech side of things and that you know doom and gloom tends to get a lot of headlines, a lot of clicks and certainly has for the last few days now. We've heard from a former Anthropic employee, politicians, and now Bridgewater's co-CIO, and that's only as of the time of recording, but you know, there's this narrative going around that we need stricter AI regulations because there's a nonzero chance that it could wipe out all of humanity in the next decade. And so, Brett, as ARC's chief futurist, what are your thoughts about that?
I think you can tell it's intentionally inflammatory because they use this idea that it's going to cause humanity to become extinct, which is just absurd and does nothing to actually solidify the argument being made. You could have in the 1850s people probably thought the limited liability corporation was going to extinct humanity because it had all profit motive and then people didn't have any of the risks and I think with every technological cycle you know you could argue well fossil fuels is going to cause humanity to go extinct because it's going to change the climate which it is changing the climate and there are ways in which like you know shifting to EVs and robo taxis is actually going to meaningfully reduce the carbon load into the atmosphere. And so every technology that's powerful, you can, you know, say there are these benefits and then these are powerful tools and they're going to cause these kind of ancillary impacts.
And I think because of the pace at which this is moving, people want to a lot of people that have power are worried about how quickly it's moving because they're worried what it means for disruption for their particular power structure. And so they are eager to glom on to a reason to try to stop it because they are being displaced. And you know I think that it's Kathy mentioned the potential kind of deflationary bust. Well deflationary busts happen because it makes debt more expensive as deflation happens. Well, for a lot of businesses that aren't in the AI supply chain because the returns on capital for building out these data centers, you know, SpaceX is going to do a 100 billion ARR or they're saying they're going to by the end of this year in their AI data center business, which really didn't exist at the beginning of this year. So, the Anthropic growth rate is amazing. The SpaceX growth rate on that particular line item is, you know, even more amazing. And that it's because the returns there are so high and that's going to pull capital out of a lot of other businesses and that naturally makes people feel anxious and then they grab onto something to try to justify why this thing shouldn't be happening which looks like it's going to disrupt what they're doing.
The fact is with disruptive technology and open-weight models in competition with closed-weight models this technology is going to move forward. I think it's likely that there will be some sort of regulatory apparatus that's built in the US. And I think it probably as and if that happens, it tends to secure the power of the companies that have navigated themselves into the lead as the kind of regulation goes in place. And so I would anticipate that's how it would occur here as well. But I think from a just a global competitiveness perspective it's important that the US kind of like regulations is relatively light touch and the companies have the closed-weight companies you know have a lot of incentive extremely high incentive to make sure that these systems are behaving as users want them to do in a way that operates within the bounds of kind of like the legal and economic frameworks that we have.
And I think that's there's actually there's this idea that an AI will be told to make as many paper clips as possible and it'll go nuts and turn the entire world into paper clips. And that's actually an out that was more likely to happen with the way that we thought AI was going to be developed 10 years ago than it is now because 10 years ago we thought well these will be purely logical rational things that you'll tell it something and it won't understand any of the context. The reality of how these tools are being created is they have they are built on human language and so they have and human ethics and human laws they have all they have too much context. They have all of the context of kind of how we operate in the world. And so just like you could say a limited liability corporation is a super intelligence. We're glomming together all the intelligences of the human and a firm to advance some notion. These are super intelligences and just like limited liability corporations. They will you know operate within the frameworks that are set out by you know humans for how to operate in the world. And so I think it's I think we'll see this cycle, we'll see other cycles of this doom and gloom. And it's partly because they are so successfully growing and delivering you know amazing productivity advances to people that people are so nervous.
Yeah, I would say I'll just add we're already we have had a glimpse at this kind of reaction from autonomous vehicles and Tesla today and Tasha knows all too well. Tesla today is really focused on safety. Why? It does appear that autonomous vehicles have crossed the threshold of human safety based on Waymo and what we believe about Tesla's data and our research. But who gets the headlines? We have I think nearly 40,000 deaths in automobiles every year. We but we don't hear much about them except for in the local papers. you know, maybe somebody we knew and so it'd be interesting for the community, but we don't hear about them nationally. We will hear about a Tesla fatality in an autonomous vehicle. And it is inevitable, too. And you could say maybe Greg Jensen at Bridgewater is not saying anything so revolutionary. But the reason it might happen is perhaps more because of the human drivers around the as autonomous vehicles are evolving into the ecosystem. So what I think about this is this is another brick in the wall of worry. And the wall of worry as we talk about in the 80s and 90s created the most durable bull market I think in history from a trend point of view and that meant two steps forward one step back but really we move forward for the most part and weathered a lot of storms back then including the S&L crisis.
Great. Thank you. Brett, the paperclip example reminds me of that Silicon Valley scene where the AI orders a pallet of beef to the office. I just connected my credit card to Muse to do some agentic shopping and test it out. So hopefully a pallet of beef doesn't arrive at my house soon, which
Or hopefully you're hungry.
Yeah. [laughter] Which so that kind of brings us into this next topic here with Frank. You know, Frank, we talk a lot about on here just the model releases and but recently there were two AI product releases, both SpaceX AI's Rockbot and Meta's Muse. Different, you know, kind of feel to these. So, could you help just explain to the listeners like I guess just basically what these are and how they differ?
Thanks, Dan. Yeah, it's very interesting. Over the last few weeks, we've seen two new AI products, Meta's Muse and SpaceX's Grockbot hit the market. And I think this is a great example of something that we've been talking about for a while now, which is that the frontier AI companies are much more than just model companies. They're increasingly agent companies and they're building these agentic products around the models that are bridging this gap between what the underlying models are capable of and how people are actually using them. And I think what Grockbot and Muse build off of is an architecture of kind of this persistent, highly connected, helpful agent that you can message very easily that we really saw the first glimpse of earlier this year with Claudebot.
What became known as OpenClaw which was a huge kind of early indicator for this type of agentic product. And if I had to describe it simply, it's bringing the kind of agentic harness that a Claude Code or a ChatGPT work have brought into the workplace and making it much lower friction to get adopted and get onboarded to the product I should say. Grockbot looks like a messaging app where you create a bot and you just tell it what you want it to do. And both Grockbot and Muse are very proactive, meaning it's not just waiting for you to send the bot a message. They can message you in advance and say, 'How can I be helpful?' Or, 'I noticed this on your calendar or you know, I can monitor your inbox and alert you if you have subscriptions that you can expire,' for example. So all of these things make it a much easier onboarding than I would say different types of agent products that we've seen so far.
And if I had to compare the two, Grockbot is probably a little bit more business forward and business focused in how it's been constructed. Whereas Muse is very clearly tuned towards the consumer which is taking a lot of Meta's experience with consumer products that serve billions of users around the world. Where it's extremely user-friendly. It comes out of the box with a ton of different connectors. So you can connect it to your email and your calendar and your financial accounts all very easily. And it gives you suggestions of different ways that it could be useful. It keeps track of to-dos. It builds a news feed for you. All of this is kind of it looks like a very full-featured consumer product and we think it's only going to get better. So we're pretty excited for this. They both offer free tiers but also subscription plans. You know, the $20 a month similar skews that we've seen from other AI companies. And for Meta in particular, I think this is a notable shift for the company, or a milestone in their AI strategy where their Muse series of models has started to perform quite well. They're not at the same capability as, I would say, OpenAI or Anthropic, but similar to SpaceX. They're kind of catching up to that frontier level. This is now a product that they can directly monetize using those models and I think they'll flex their distribution muscles to get that product out into the market and see how consumers react to it. So we're really excited to see these two products and really agents making their way into the hands of more everyday users.
Thank you, Frank. And I have one more for you. Staying on the AI side here. So speed has speed of like inference and speed of response has kind of turned into its own axis in the AI world and labs are designing their own inference chips and specialists are pushing token throughputs past what GPUs can deliver. So why does speed matter so much and who captures the value here?
Happy to talk about this because I think it's a really interesting evolution in the market for AI tokens. We've seen companies like Grock which was essentially acquired by Nvidia back at the end of last year and then now Cerebras which recently IPOed and is held across our relevant strategies both competing in this space and offering these ultra-fast inference services. So to give you an example Cerebras their chip and their system CS3 can run OpenAI's 5.6 Soul model. This is their latest model outside of the hot off the press Astra model. It can run 14 times faster for a user than the standard model that's running on predominantly Nvidia chips. So what that means and this is a real example that I had recently when you're working on a use case with an agent maybe it's building a slide deck maybe it's building a software application now that the models can accomplish longer time horizon tasks and do much more work for you autonomously oftentimes you're waiting for that model tens of minutes. So I had a task I was working on that took 20 minutes. With the partnership between Cerebras and OpenAI. They're currently in early access beta with this, but Cerebras running 5.6 Soul can run 14 times faster. So that means something that took 20 minutes now can take less than a minute and a half. So the interactivity and the workflow can help you stay kind of in that flow state iterating with the model much more quickly. You get more intelligence per unit of time so to speak and I think that is a growing need in the market as you're seeing models work for longer time periods but there's other reasons why you would want to have fast inference.
Another example I think is voice the latest voice models from OpenAI and Anthropic they can they can stop to think before they respond which results in a much higher answer quality we saw actually with the release of reasoning models originally in the text world. Now that capability is coming to the voice world but if you're thinking of a real-time voice conversation it's not a great experience if a model stops and says nothing for a minute while it's thinking about your answer. So if you can speed that up by 10 14 15 times you can result you can deliver a very interactive voice experience without having that latency that's inherent with introducing thinking into the process. The last example is just use cases that are very time-sensitive where you're willing to pay up for speed. One of the early access customers that's demoing 5.6 Soul ultra-fast is Jane Street. And you can imagine how a company in the quant trading space would want to use the smartest intelligence and they want to be able to deploy it as quickly as possible so they can capture opportunities in the market faster than their competition. So those are all the use cases for it.
Economics I think are also interesting. Because of these use cases there's a willingness to pay up for speed. So OpenAI and Anthropic already offer fast modes for their APIs where for the 5.6 models, for example, you can pay twice as much per token to get twice the speed. So there's this demonstrated willingness to pay. I find myself turning fast mode on quite a bit. But ultra-fast mode, which rather than two times faster, will be 14 times faster. We expect companies like OpenAI that offer this service powered by chips that Cerebras offers will charge at a premium even higher to that. And for all the reasons I mentioned before, we think there's going to be a strong willingness to pay for that. This whole market paradigm I think is relatively new because these fast inference chips are just starting to hit the market from Cerebras for one and then also Nvidia announced Grock powered racks recently as well. So we're excited to see how this market evolves and happy to have exposure to companies like both Cerebras and Nvidia in our portfolios.
Great. Thank you, Frank. We'll move into the venture side of things here. Brett, I have one for you. Just could you please highlight a few of the new companies that have found their way into the ARC venture fund, the ARKVX ticker fund?
