"Eventually we'll become a brand name”

It took eight months of negotiations, but it’s happening - Infineon Technologies (NYSE, XETRA: IFX), the semiconductor spin-off from Siemens (NYSE: SI; XETRA: SIE), and Israeli start-up Saifun Semiconductors, are taking their cooperation a step further.

Ingentix, the joint Siemens-Saifun venture announced in May 2001, will be enlarged to handle additional technologies. This will enable Infineon to offer more memory products, while Saifun will be able to provide technology for a growing market, which is expected to reach $15 billion in the next five years. In addition, the ownership of the joint venture has changed. Infineon will be a 70% partner, and Saifun 30%, compared with 51% for Infineon and 49% for Saifun under the previous arrangement. The new partnership division will not affect Infineon’s direct holdings in Saifun itself.

The joint venture will henceforth be called Infineon Flash. The two partners are trying to combine Saifun’s NROM non-volatile memory, which does not require electric power to preserve information, with Infineon’s micro-controllers. They plan to expand beyond the mobile flash memory cards developed by Ingentix. Another goal is to obtain a foothold in the field of stationary memory used in cell phones and smart cards. Infineon memory chip division head Dr. Harald Eggers explains that Infineon Flash wants to position itself as a supplier of all types of memory, with the help of technology developed by both partners, instead of only supplying DRAM memory, as at present.

Infineon’s financial statements for 2001, the year the original joint venture was set up, showed the company was planning to invest $17 million in Ingentix. Eggers said Infineon was not planning to inject further capital into Infineon Flash, since most of the necessary investments in the venture had already been made, including a $5 million manufacturing facility in Dresden. “As far as the anticipated volume of activity is concerned, we’re fairly confident that the existing infrastructure will suffice for the joint venture. In another three years, we may have to reconsider the question of investment in manufacturing, but until then, we’re in good shape,” Egger added.

In any case, the company has agreements to manufacture at the SMIC plant in China, and the Nanya facility in Taiwan. Although Eggers asserts that the venture does not need these agreements to expand its manufacturing capacity, they may come into play later on. “We’re in effect preparing a reserve production capacity for the joint venture for the coming years, which we will use as needed.”

Infineon’s memory division lost money in 2001-2002, but Eggers says it will make a profit this year. Several months ago, the division heads announced their goal of becoming one of the world’s four leading semiconductor suppliers. Infineon’s current initiative follows several operations in 2002, which were designed to consolidate existing markets in the memory field, and penetrate new ones.

The 2002 operations included an agreement with Taiwanese company Windbond to jointly purchase and share know-how for producing DRAM chips from 300 mm wafers, an agreement with Nanya to use 0.9 and 0.7 micron technologies to develop technology for producing DRAM chips from 300 mm wafers, and a partnership with Advanced Micro Devices (Nasdaq: AMD) and Dupont’s (NYSE: DD) masking division to develop masking technology for producing the next generation of semiconductors.

Although over 18 months have passed since Infineon and Saifun announced the founding of Ingentix, the latter’s memory cards are still not on the market. Eggers and Saifun founder, president, and CEO Boaz Eitan claim that samples of 64 megabyte (MB) cards have already been given to potential customers for testing, but under a different model. Instead of selling the technology and cards through OEM agreements, the company will sell the cards by itself. The company hopes to issue its first product, a 512 MB memory, together with two others: a memory that can be embedded in an electronic device, and a mobile memory multimedia card (MMC) containing the memory. Infineon hopes to compensate for being late to market, where there are already similar cards of over one gigabyte (GB), by including security on the card, and giving what it claims is superior performance.

At the 2003 International CES exhibition, Sony (NYSE: SNE) and Panasonic announced they would soon produce a 2 GB flash memory card. Eitan is not deterred by the competition, claiming that his company’s current technology already supports this amount of memory on flash memory cards. He adds that the company plans to display cards this year with various memory capacities, both greater and less than 512 MB.

In June 2002, SanDisk Corporation (Nasdaq: SNDK), Matsushita Electric Industrial (NYSE: MC), Toshiba, and Ingentix announced they had formed a consortium to establish a joint standard for flash media card security. The companies had hoped to present the standard by October 2002, but the standard still seems far from completion.

According to Eitan, the current card will not support the joint standard that the companies are trying to establish. He claims, however, that the five companies will in any case adopt an Infineon security standard for using smart cards in commercial electronic applications (smart digital wallets, G.N.). As far as Eitan is concerned, Infineon’s flash cards are well protected. “It will be a while before we see cards including a joint security standard,” he predicts. “Security on today’s flash memory cards is minimal, and relates to matters such as copyright. It’s harder to secure financial applications; that’s no trivial problem. That’s what the consortium is trying to do.” Eggers: ”Infineon has a 39% share of the smart cards market, and we have quite a few years of experience in providing security technology for memory cards. I believe that this know-how will also fit in with the joint venture, and enable these companies to use their experience to achieve a joint standard.”

Infineon current sells its smart cards to financial institutions, credit card companies, and medical institutions. Infineon expects to utilize its current marketing channels to sell the smart cards, which will include memory components from Infineon Flash. The memory cards will be sold to Asian companies, such as Kingston, PQI, and Axsun Technologies (and probably M-Systems Flash Disk Pioneers (Nasdaq: FLSH), which has invested in Saifun). Other potential customers come from various computer and communications fields: manufacturers of routers, handheld devices, and cellular telephones.

The memory market is divided into different fields; consequently, rating the players in this market often leads to distortions. Certain companies, such as Intel (Nasdaq: INTC) have a substantial market share in one field, but are completely absent from another field. In the stationary flash area, (PCs, cellular telephones, and handheld devices) Infineon will go head to head against market leader Intel, which earned more than $24 billion from its memory business in 2002. Far behind are such competitors as Samsung, STMicroelectronics (NYSE: STM), Texas Instruments (NYSE: TXN), AMD, and Toshiba.

In the market for flash memory cards (used to store media on digital cameras, music players, handheld devices, and cellular telephones), Infineon expects to confront Samsung, SanDisk, Toshiba, and Sony, among others. In the DRAM memory field, Infineon is still in fourth place, after Samsung, Micron Technology (NYSE: MU), and Hynix Semiconductor (KSC: 00660.KS), but managed to boost its market share to 12.2% in 2002, compared with 9.7% in 2001. Samsung, the leader, increased its market share form 27% to 30.8%.

AMD and Fujitsu’s products will compete again those of Infineon next year, and both will have Saifun’s technology. This situation does not seem particularly strange, for those following this industry. SanDisk itself sells flash memory cards that compete with Sony’s memory stick products, but also provides Sony with the technology for its next generation of memory cards. “No company is willing to work with only supplier. Cooperation between competitors is good for all the suppliers,” Eggers explains.

Eggers prefers not to give precise forecasts for the companies’ revenue from the joint venture. He notes, however, the flash memory market will amount to an estimated $15 billion in the next five years. He says Infineon has a 14% share of the DRAM memory market, and plans to achieve a similar share in the other memory sectors.

Eitan states that the company already has substantial revenue from companies to which it has licensed its technology. “Our initial strategy was to sign license agreements for our technology. Our strategy has changed to joint ventures, but we’re not neglecting the license sale model. The joint venture agreement with Infineon will be the last we’ll sign for flash memory cards and storage memory components. On the other hand, we’re very interested in signing agreements for embedded memory with companies manufacturing logical memory. We recently signed such an agreement, and I believe more are on the way.”

Candidates for such agreements include Microchip Technology (Nasdaq: MCHP), which markets controllers that include memory; Bluetooth component manufacturers, which need this type of memory to activate the features of their components; and even the relevant division in Saifun’s old partner, Infineon. “I assume some of the embedded memory agreements will be signed with our customers in other areas, will others will be signed with new customers,” Eitan says. “License deals have been very successful for us, and will continue to be. We don’t plan to hire many people to sell our products under the Saifun brand name, because in this industry, it takes a lot of time and effort to build such a brand. Cooperation and technology license sales make it possible to earn substantial revenue much more quickly, and with much less effort. We’ll eventually build our own brand name, but we don’t plan to reach this stage for some years yet.”

Regarding the company’s cash flow, Eitan says Saifun’s bank account is growing every year, thanks to license sale agreements and proper expense planning. Saifun, which has raised $77 million to date, has no need to raise more capital, except to build its own production facility, which right now seems far in the future, while its license sale and cooperation strategy persists. The only figure Eitan is willing to disclose is that Saifun doubled its cash flow from 2001 to 2002, and will double its bank account in the next 18 months.

Eitan declares that neither Saifun, nor its joint venture with Infineon, will hold a public issue, unless capital is required to construct a manufacturing plant. Furthermore, it will be necessary to wait for a market recovery. “Right now, we’re busy with current projects,” Eitan says. “We’ve no time right now to dream about future projects.”

One bits, four bits

The technology used in the products made possible by Saifun is based on the capacity to store two bits of information in each cell, compared with only one bit in the previous generation of memory components. Doubling the cell’s storage capacity makes it possible to either double the storage volume of each memory component, or halve the storage area, while retaining the same storage volume. Documents attesting to Saifun’s 75 patents related to its various memory technologies adorn the corridors of the company offices in Netanya.

Most of the media hype about Saifun, however, has ignored the abilities of its competitors, some of which are linked to Saifun in various ways. AMD, of course, has MirrorBit technology, which Saifuns claims was copied from its own technology. A lawsuit ended with Saifun getting a huge contract and investment from AMD and Fujitsu. The other known technologies for putting two bits in one cell, called multi-level cell (MLC), use changes in the electric current fed to each transistor.

In the conventional storage method, every cell contains a 0 or 1. The difference between the two values lies in the way the transistor interprets the electric current below a certain level as 1, and above a certain level as a command to erase the information, leaving a 0. MLC technologies make is possible to read up to four different ranges of electric current directed at the transistor, and hence, in principle, to store four bits in one cell.

Most large companies in the memory market already use this storage method, but as of now, only two bits are stored in a cell. Intel has a technology called StrataFlash, which STMicroelectronics has obtained a license to use, and which Toshiba and SanDisk are already using. Samsung has signed a patent-sharing agreement with SanDisk, under which Samsung gained access to SanDisk’s MLC know-how. We’ll see the first developments based on four bit per cell storage next year.

Incidentally, Toshiba and M-Systems displayed a product combining Toshiba’s MLC-based memory with M-Systems’ x2 technology the latest version of M-Systems’ DiskOnChip. M-Systems asserts that the inclusion of x2 solves MLC’s slowness and loss of information problems.

In addition, miniaturization has affected the companies’ manufacturing: from 0.18 microns, to 0.13, 0.11, and even 0.09. This has made it possible to produce more chips from each silicon wafer, and has also made the components themselves smaller, even without writing more information on each component.

”Globes”: Have you lost your relative advantage?

Eitan: ”Storing four bits per cell is something that others are incapable of doing, and I stand by that. In order to store four bits per cell, they’ll have to store 16 separate information layers on top of each other. In physical terms, the signal isn’t strong enough, compared with the noise created by the density, to make this possible without losing information. Their methods may be practical in digital format music storage, which is more tolerant of information loss in small groups, but not with applications in which the loss of information is critical, such as communications and security.”

Published by Globes [online] - www.globes.co.il - on February 25, 2003