Using AI to Pick AI Stocks

By Hamlin Lovell, NordicInvestor

The fund ODDO BHF Artificial Intelligence uses AI to pick stocks, investing in companies that are taking advantage of AI. It is differentiated from other technology- oriented strategies and funds by a more dynamic and diversified approach, including more small and mid-caps and less large caps.

“The fund aims to identify companies positively exposed to the AI thematic by combining three complementary investment steps: a semantic & sentiment analysis on unstructured data, a multi factor quantitative model and the human review of the companies selected. This innovative and unique approach is clearly differentiating us from fundamental stock picking funds that will often apply only one of these pillars”, says Maxence Radjabi, who co-manages the strategy together with Brice Prunas, both fund managers at ODDO BHF Asset Management SAS.

ODDO BHF Artificial Intelligence also looks beyond the obvious companies in the technology sector to exploit the benefits of AI. “Tech companies were the enablers of new technology, and they enjoyed a lot of growth, according to our analysis. Now new and disruptive companies in other sectors are also using AI to gain competitive advantage and market share. For instance, new oncology treatments for cancer and minimally invasive surgery have created revolutions in their own sectors. We have a more diversified approach to the theme than purely tech sector-based funds”.

Limited exposure to large cap tech stocks differentiates ODDO BHF Artificial Intelligence from market cap weighted technology indices and strategies. Currently, the fund ODDO BHF Artificial Intelligence owns four of the FANGMAN (Facebook*, Amazon*, Netflix*, Google/Alphabet*, Microsoft*, Apple*, Nvidia*) complex and one of the BAT complex in China (Baidu*, Alibaba*, Tencent*). The holdings change with quarterly rebalancing. Given the fact that the 60 stocks are equally weighted, the maximum exposure to FANGMAN and BAT combined would be just under 17% if the fund was exposed to all ten stocks; and it has never gone over 10% so far

Key themes: big data, automatization, machine learning, deep learning, and cognitive computing

Though stock positions are equally weighted, the weighted average allocations to the five big themes of big data, automatization, machine learning, deep learning, and cognitive computing need not be equally weighted. In any case, those themes are only the top five as of 2019. They are a moving target. “We expect to identify new trends and sub-trends in future. For instance, big data might get smaller and new themes may emerge”.

Currently, the fund has identified eight sub-sectors that are exploiting these themes: autonomous vehicles (AVs); Advanced Robotics; Internet of Things (IoT); Automation; NLP; Machine or Computer Vision; Virtual agents or assistants, and cloud technology. These are clearly at different stages of maturity “Cloud storage, already growing at 40-50% per year, is probably at the most advanced stage of development, though a lot of the growth may already be priced in. In contrast, computer vision in areas such as autonomous vehicles, factory automation, or speech recognition and analysis, have still enormous potential in the future”, says Radjabi.

“Our holdings include Japanese mid cap Advantest*, which shares a duopoly on automated semiconductor testing with Teradyne. CarGurus* is an example of how e-commerce is transforming the automotive industry. By leveraging AI, it is changing the relationship between consumers and car dealers”, says Radjabi.

Russia’s Yandex* effectively has two near monopolies internet search, and cab ride hailing. The fund’s emerging market exposure is in practice limited. “Although China is the second biggest AI player in the world, it has only a few listed AI companies beyond the BATs. However, we are exposed to China through insurance company Ping An, an example of how AI will transform other sectors. Our direct exposure to EM listed firms is probably only 5-10%, though the share of portfolio companies’ revenues coming from EM is probably higher”, he adds.

Active share is above 90%, in common with ODDO BHF AM’s other equity funds, but the AI fund has less than 5% overlap with the firm’s other funds, partly because ODDO BHF AM is renowned for European equities, where fewer innovative companies are found. “Europe is clearly lagging behind on publicly listed AI stocks, though there are some semiconductor and hardware makers. We tend to invest in more diversified firms leveraging AI. In the Medical Technology (MedTech) sector we have owned Sweden’s Elekta*, which has developed ways of using IBM Watson and radiology for cancer treatments, but sold it this summer and switched into US competitor Varian*”, says Radjabi.

A unique investment process

One exception to the general lack of European exposure is the company that helps ODDO BHF AM with the analytical process used to select stocks. ODDO BHF AM has formed an exclusive partnership with a French leader in data science, which provides the AI and NLP techniques and algorithms used for the first systematic part of the investment process: the analysis of the global market news flow  – four million sentiment and semantic data points per day – to identify subthemes and a thematic index of 300 companies.

In the second step, the quantitative model “Algo 4” screens stocks based on four factors (valuation, quality, momentum and market capitalization). “Algo 4” has been successfully applied by ODDO BHF AM for more than 10 years in its quantitative equity management. This reduces the universe of 300 AI stocks down to 60 equities exhibiting the most interesting financial and risk profile according to ODDO BHF AM. “The model has been fine-tuned and adapted for use in a thematic and global strategy. The purpose of the strategy is to provide an exposure on this key theme that we hope will outperform the global Equity market over the long term”.

ODDO BHF AM’s own analytics focus on the financial health of a company including the three key pillars of the approach: growth and quality and value metrics. Some technology companies, including privately owned “Unicorns”, are unprofitable – and some of them are still losing money after their IPOs. “In September and October 2019 there has been some rotation away from highly valued firms, and recent IPOs such as Workday* and Atlassian* have seen share price drops despite maintaining earnings guidance. This suggests that embedded expectations US cloud software companies like Workday or Atlassian must have been very high. When sentiment changes, the derating can be violent and fast”, he says.

Finally, the portfolio is subject to a qualitative review conducted by the fund managers. The purpose of this review is to identify any potential incoherence in the portfolio and to deepen the fundamental insight into the selected companies, in particular through meetings between the companies and the managers. 

Value and Growth

The average valuation of the fund ODDO BHF Artificial Intelligence is a PE ratio of 18.9 times, which is near identical to the MSCI World NR on 18.7 times. “This is coincidental as there is no specific target to maintain identical valuations. And the average belies a wide range: some firms owned are on PE ratios above 19 while others, for instance, in hardware and semiconductors, are on much lower valuations”, says Radjabi.

The portfolio does not in fact have the most spectacular earnings growth forecast. This is partly complicated by the fact that semiconductors saw a cyclical pullback in earnings. Overall, the portfolio is a balance between pure play AI stocks with very high growth and more diversified firms leveraging AI to improve their growth, gain market share or even achieve a turnaround in their business. 

*Those stocks are no investment recommendations

2019-11-08T13:19:09+00:00By |Categories: Equities, The Nordic Brief|