Upstart: A Fintech Stock To Buy

This is a recent article that I wrote on Upstart that just posted today. I include a major risk that upstart has. I titled the article Upstart: A Fintech Stock To Buy On The Pullback:


Upstart is in the midst of a major pullback since analysts began downgrading the stock in mid-October and the stock price hasn’t recovered after its most recent earnings release.

In the most recent earnings report, the company is still showing triple digit year-over-year revenue growth, while at the same time being both profitable and Free Cash Flow Positive.

If Upstart continues its success in showing higher approval rates, lower defaults, and lower loan payments for consumers compared to the FICO score, the stock will be a market-beater.

The company has very high risk, as innovation with such things like Artificial Intelligence has not fared well in the financial industry in the past.

Upstart is a buy on this most recent pullback.

Read More:…



As per usual I loved the this article you wrote, Starrob.

Regarding what you said here:
Investors will need to watch this company closely, especially during times of rising interest rates or during times when recession is likely, which are environments which Upstart’s AI models have never seen before. Whether Upstart’s AI models can handle poor credit environments remains to be seen but it is almost certain that this stock would likely undergo a severe downdraft, in such an environment.

Thank you for looking at another possible downside risk here.

As I and others stated, clearly Upstarts actual trajectory of growth has changed. However, perhaps unlike those that sold out, I believe that 1-3 years from now when we go back and draw the line of best fit demonstrating the growth of Upstart, the trajectory of this line will be better than any outside of those with higher allocation in my portfolio :grin:. Along with the incredible advantages afforded to first movers in AI, I’m willing to face the fact that the magic of proprietary AI is done in the proverbial black box and therefore we are unable to pin point exactly how they accomplish what they do; as we are unable to predict when the tech will make some sort of leap forward. Unless we expect Upstart to regularly pre announce before scheduled earnings, more than any other type of company perhaps we must trust in management and the engineers tweaking the algorithms. My albeit limited understanding prior to Earnings was that we’d already saw what would happen to upstarts revenue generation during a general economic downturn during the Pandemic. And the only downside surprises to future guidance would be similar to what we experienced this last quarter (save for a IMO highly unlikely reputational/existential breach) and there is a whole lot of room for upside surprise.

I really appreciate your look at the risks around rising interest rates/Inflation risk. Over the last few yearsmIvemtaken Saul’s advice to ignore macro/geopolitical risk when investing. But, I have to say that the overall Market is likely going to look at this considering Upstart is considered a fintech company by most.

If I understand your article correctly you agree, the actual path taken to reach this incredible level of growth for which I expect Upstart to achieve in 1-3 years may show greater volatility than any other company in my entire portfolio. We’re not going to try and time when Upstart will make an unexpected leap in growth, from adjusting their algorithms or rolling out services unexpectedly faster than they predicted. And we’re not going to try to predict, nor do I expect Upstart Leadership to predict, when hackers require Upstart to divert resources and perhaps stop approving loans for a while🤕. (Note: As far as we know, this recent attempt to hack Upstart merely required Upstart to dig themselves a deeper moat when they figured out how to resolve this issue). GeoPolitical or MacroEconomic effects on Upstart are beyond this companies control; however IMO, if any fintech was able to usefully take these risks into consideration and quickly adjust course accordingly, I believe it would Upstart. In fact they have said as much by my estimation of their commentary during the Conference Calls I’ve heard. These unpredictable issues may again lead to a disappointing beat, as was the case with this last quarter numbers. I’m not hoping it never happens again. I’m planning on it.

Sincere thanks for making this investment experiment more interesting,



Hey Rob,

This was an interesting read and a very nice analysis of the potentials and risks in Upstart. I was intrigued about the story of First Marblehead. The natural question that arises, is “Is this time different” and, “if so, why?”.

I am not an expert in AI, but I know it is progressing far faster than we could have imagined. I had coffee with an AI researcher in 2010 who described to me the problem of creating an AI that could beat a human at Go!. He said, it might not ever happen, and if it did, it was many years down the road. When AlphaGo beat the world Go! champion in 2016, I was stunned. I knew it was a huge achievement, and it had happened much faster than professionals in the field had expected.

And that was five years ago…

Saul has lost his voice saying that this is not a tech bubble. Everyone remembers 2000 and realizing that “this time is different” has helped people to earn a fortune.

We may be entering the “this time is different” phase of AI. People remember past AI failures, not to mention current AI failures (Zillow). The result is that there will be enduring skepticism even towards enduring winners in AI.

The analysis on this board has me confident that Upstart is a winner. It may not be as easy to project as some other AI stocks who also have a SaaS model. But the results speak for themselves. All I see is continuing hyper growth, solid leadership, and a product that is both successful and highly needed. I like how you point out in your article that Upstart contributes to ESG goals of financial inclusion.

Great article. Thanks again.


We may be entering the “this time is different” phase of AI. People remember past AI failures, not to mention current AI failures (Zillow). The result is that there will be enduring skepticism even towards enduring winners in AI.

Just MHO, but the first sentence can be dangerous thinking. We are certainly learning more about how to create an AI, and better at creating the environment in which an AI can run at speed so the results can be applied in an appropriate business timeline. But AI is only ever going to be as good as the models, data, and analysis that underly the target, and those are all designed by humans (at least for now). People and businesses will continue to create AI failures, for any variety of reasons: not fully understanding the target and failing to account for all the parameters to feed into the AI algorithms; not understanding the data they are receiving about the target, or missing what the data is saying; not identifying the correct target in the first place.

Each target for an AI project is going to have a particular set of parameters that are critical to understanding it, and each parameter may require further analysis to determine how to break it down. Take age for example. According to Malcolm Gladwell’s Outliers book*, the odds of a kid becoming a hockey player are higher if they are born earlier in a year than later, because, on average, they are that much bigger and more physically capable than their age mates born later in the year. They are more capable, and therefore get more attention, from coaching to ice time.

As it relates to Upstart, could something similar be a predictor of how likely someone would be to pay off a loan? Are age groups of 5-10 years good enough to keep the algorithm relevant, or does birth month actually matter? Maybe kids born later in the year aren’t playing hockey (because they can’t compete), but studying instead. You can’t really know until you test it, and you can’t test it if you never thought to ask the question. The AI isn’t magic, it won’t know to even consider age unless you make it part of the data set and tell it which aspects of “age” to look at.

tldr; we will probably never be in a “this time is different” phase of AI. Caution and demonstrated results are all we can go on.

  • there is some debate about the book and its claims (it may apply better to junior hockey than the NHL if at all), but it’s still a suitable example of the kind of thinking that needs to go into designing an AI.

The underlying strength of UPST rests on two things: (a) increasing the pool of potential borrowers; (b) doing a better job than FICO in predicting risk.

“Jack’s” FICO score is reactive and based on his individual consumer behavior. It matters not to FICO if Jack lost his job two months ago after having six months of reduced hours & wages. As long as Jack’s paying the bills, all’s good with FICO

UPST models are predictive and based upon a multitude of additional factors. The data they collect may show that Jack is meeting his financial obligations, but that he is on a major financial precipice – no longer in one of the data model’s clusters of ‘desirable customers’.

UPST has proven it can beat FICO in current conditions. Because of the above, it is my opinion that it will beat FICO by an even wider margin in economic downturns.

[Remaining long on UPST]